|
View:
New views
20 Messages
—
Rating Filter:
Alert me
|
| < Prev | 1 - 2 | Next > |
|
|
mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Hi,
We're running Nexus 1.3.6 on a separate linux box on our intranet. During 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn prompt sits idle for 3 minutes, eventually carrying on w/o an error. During those 3-4 minutes: - viewing that repo in a browser does NOT list the new WAR file - 'ls -la' on hosted repo's directory shows something slowly writing out myartifact.war.tmp. Shortly after the .tmp file is fully written (== in size to the war), the actual WAR is created and web listing of the repo is updated. The repo's configured with "Include in Search = false" - so I am assuming indexing should not be happening. Does Nexus do some sort of other processing on the WAR before making it available for download? If so, any way to disable that (and, thus, free devs from having to wait for 3 min while mvn deploy's running). For comparison, scp'ing the same file between from the same dev machine to the Nexus box took < 19s. thanks! -nikita ----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Sounds exactly like the behaviour we had with
https://issues.sonatype.org/browse/NEXUS-2879 which we have encountered when our nexus has been running for a while. Restarting nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we get a chance we are planning to upgrade. Until then the issue has not repeated for us (I think it requires a long uptime with lots of big deployments -Stephen 2009/11/9 dukehoops <nikita.tovstoles@...>: > > Hi, > > We're running Nexus 1.3.6 on a separate linux box on our intranet. During > 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn prompt > sits idle for 3 minutes, eventually carrying on w/o an error. During those > 3-4 minutes: > - viewing that repo in a browser does NOT list the new WAR file > - 'ls -la' on hosted repo's directory shows something slowly writing out > myartifact.war.tmp. Shortly after the .tmp file is fully written (== in size > to the war), the actual WAR is created and web listing of the repo is > updated. > > The repo's configured with "Include in Search = false" - so I am assuming > indexing should not be happening. Does Nexus do some sort of other > processing on the WAR before making it available for download? If so, any > way to disable that (and, thus, free devs from having to wait for 3 min > while mvn deploy's running). > > For comparison, scp'ing the same file between from the same dev machine to > the Nexus box took < 19s. > > thanks! > > -nikita > > ----- > ---------------- > Nikita Tovstoles > vside.com > ---------------- > > -- > View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html > Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?These don't sound related. Specifically in Nexus 1.3.6+, the files are
first written to the .tmp location as they stream in, then it's renamed to the final location. If you are seeing it slowly come in, this tells me it's not the indexing, but something between you and Nexus. Are you running Nexus on a VM or behind some reverse proxy? On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly <stephen.alan.connolly@...> wrote: > Sounds exactly like the behaviour we had with > https://issues.sonatype.org/browse/NEXUS-2879 which we have > encountered when our nexus has been running for a while. Restarting > nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we > get a chance we are planning to upgrade. Until then the issue has not > repeated for us (I think it requires a long uptime with lots of big > deployments > > -Stephen > > 2009/11/9 dukehoops <nikita.tovstoles@...>: >> >> Hi, >> >> We're running Nexus 1.3.6 on a separate linux box on our intranet. During >> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn prompt >> sits idle for 3 minutes, eventually carrying on w/o an error. During those >> 3-4 minutes: >> - viewing that repo in a browser does NOT list the new WAR file >> - 'ls -la' on hosted repo's directory shows something slowly writing out >> myartifact.war.tmp. Shortly after the .tmp file is fully written (== in size >> to the war), the actual WAR is created and web listing of the repo is >> updated. >> >> The repo's configured with "Include in Search = false" - so I am assuming >> indexing should not be happening. Does Nexus do some sort of other >> processing on the WAR before making it available for download? If so, any >> way to disable that (and, thus, free devs from having to wait for 3 min >> while mvn deploy's running). >> >> For comparison, scp'ing the same file between from the same dev machine to >> the Nexus box took < 19s. >> >> thanks! >> >> -nikita >> >> ----- >> ---------------- >> Nikita Tovstoles >> vside.com >> ---------------- >> >> -- >> View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >> Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. >> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Well we were getting the upload of our 100MB+ zip file in < 5 seconds
and then the mvn prompt sitting idle for 3-4 minutes before continuing w/o error. It was only when we went to a bigger zip file that we got the error. So that sounds like the same thing! ;-) -Stephen 2009/11/10 Brian Fox <brianf@...>: > These don't sound related. Specifically in Nexus 1.3.6+, the files are > first written to the .tmp location as they stream in, then it's > renamed to the final location. If you are seeing it slowly come in, > this tells me it's not the indexing, but something between you and > Nexus. Are you running Nexus on a VM or behind some reverse proxy? > > On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly > <stephen.alan.connolly@...> wrote: >> Sounds exactly like the behaviour we had with >> https://issues.sonatype.org/browse/NEXUS-2879 which we have >> encountered when our nexus has been running for a while. Restarting >> nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we >> get a chance we are planning to upgrade. Until then the issue has not >> repeated for us (I think it requires a long uptime with lots of big >> deployments >> >> -Stephen >> >> 2009/11/9 dukehoops <nikita.tovstoles@...>: >>> >>> Hi, >>> >>> We're running Nexus 1.3.6 on a separate linux box on our intranet. During >>> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn prompt >>> sits idle for 3 minutes, eventually carrying on w/o an error. During those >>> 3-4 minutes: >>> - viewing that repo in a browser does NOT list the new WAR file >>> - 'ls -la' on hosted repo's directory shows something slowly writing out >>> myartifact.war.tmp. Shortly after the .tmp file is fully written (== in size >>> to the war), the actual WAR is created and web listing of the repo is >>> updated. >>> >>> The repo's configured with "Include in Search = false" - so I am assuming >>> indexing should not be happening. Does Nexus do some sort of other >>> processing on the WAR before making it available for download? If so, any >>> way to disable that (and, thus, free devs from having to wait for 3 min >>> while mvn deploy's running). >>> >>> For comparison, scp'ing the same file between from the same dev machine to >>> the Nexus box took < 19s. >>> >>> thanks! >>> >>> -nikita >>> >>> ----- >>> ---------------- >>> Nikita Tovstoles >>> vside.com >>> ---------------- >>> >>> -- >>> View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >>> Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. >>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>> For additional commands, e-mail: nexus-user-help@... >>> >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?What version of Maven?
On Tue, Nov 10, 2009 at 3:47 AM, Stephen Connolly <stephen.alan.connolly@...> wrote: > Well we were getting the upload of our 100MB+ zip file in < 5 seconds > and then the mvn prompt sitting idle for 3-4 minutes before continuing > w/o error. It was only when we went to a bigger zip file that we got > the error. > > So that sounds like the same thing! ;-) > > -Stephen > > 2009/11/10 Brian Fox <brianf@...>: >> These don't sound related. Specifically in Nexus 1.3.6+, the files are >> first written to the .tmp location as they stream in, then it's >> renamed to the final location. If you are seeing it slowly come in, >> this tells me it's not the indexing, but something between you and >> Nexus. Are you running Nexus on a VM or behind some reverse proxy? >> >> On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly >> <stephen.alan.connolly@...> wrote: >>> Sounds exactly like the behaviour we had with >>> https://issues.sonatype.org/browse/NEXUS-2879 which we have >>> encountered when our nexus has been running for a while. Restarting >>> nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we >>> get a chance we are planning to upgrade. Until then the issue has not >>> repeated for us (I think it requires a long uptime with lots of big >>> deployments >>> >>> -Stephen >>> >>> 2009/11/9 dukehoops <nikita.tovstoles@...>: >>>> >>>> Hi, >>>> >>>> We're running Nexus 1.3.6 on a separate linux box on our intranet. During >>>> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn prompt >>>> sits idle for 3 minutes, eventually carrying on w/o an error. During those >>>> 3-4 minutes: >>>> - viewing that repo in a browser does NOT list the new WAR file >>>> - 'ls -la' on hosted repo's directory shows something slowly writing out >>>> myartifact.war.tmp. Shortly after the .tmp file is fully written (== in size >>>> to the war), the actual WAR is created and web listing of the repo is >>>> updated. >>>> >>>> The repo's configured with "Include in Search = false" - so I am assuming >>>> indexing should not be happening. Does Nexus do some sort of other >>>> processing on the WAR before making it available for download? If so, any >>>> way to disable that (and, thus, free devs from having to wait for 3 min >>>> while mvn deploy's running). >>>> >>>> For comparison, scp'ing the same file between from the same dev machine to >>>> the Nexus box took < 19s. >>>> >>>> thanks! >>>> >>>> -nikita >>>> >>>> ----- >>>> ---------------- >>>> Nikita Tovstoles >>>> vside.com >>>> ---------------- >>>> >>>> -- >>>> View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >>>> Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. >>>> >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>> For additional commands, e-mail: nexus-user-help@... >>>> >>>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>> For additional commands, e-mail: nexus-user-help@... >>> >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?2.0.9, 2.0.10, 2.2.1
2009/11/10 Brian Fox <brianf@...>: > What version of Maven? > > On Tue, Nov 10, 2009 at 3:47 AM, Stephen Connolly > <stephen.alan.connolly@...> wrote: >> Well we were getting the upload of our 100MB+ zip file in < 5 seconds >> and then the mvn prompt sitting idle for 3-4 minutes before continuing >> w/o error. It was only when we went to a bigger zip file that we got >> the error. >> >> So that sounds like the same thing! ;-) >> >> -Stephen >> >> 2009/11/10 Brian Fox <brianf@...>: >>> These don't sound related. Specifically in Nexus 1.3.6+, the files are >>> first written to the .tmp location as they stream in, then it's >>> renamed to the final location. If you are seeing it slowly come in, >>> this tells me it's not the indexing, but something between you and >>> Nexus. Are you running Nexus on a VM or behind some reverse proxy? >>> >>> On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly >>> <stephen.alan.connolly@...> wrote: >>>> Sounds exactly like the behaviour we had with >>>> https://issues.sonatype.org/browse/NEXUS-2879 which we have >>>> encountered when our nexus has been running for a while. Restarting >>>> nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we >>>> get a chance we are planning to upgrade. Until then the issue has not >>>> repeated for us (I think it requires a long uptime with lots of big >>>> deployments >>>> >>>> -Stephen >>>> >>>> 2009/11/9 dukehoops <nikita.tovstoles@...>: >>>>> >>>>> Hi, >>>>> >>>>> We're running Nexus 1.3.6 on a separate linux box on our intranet. During >>>>> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn prompt >>>>> sits idle for 3 minutes, eventually carrying on w/o an error. During those >>>>> 3-4 minutes: >>>>> - viewing that repo in a browser does NOT list the new WAR file >>>>> - 'ls -la' on hosted repo's directory shows something slowly writing out >>>>> myartifact.war.tmp. Shortly after the .tmp file is fully written (== in size >>>>> to the war), the actual WAR is created and web listing of the repo is >>>>> updated. >>>>> >>>>> The repo's configured with "Include in Search = false" - so I am assuming >>>>> indexing should not be happening. Does Nexus do some sort of other >>>>> processing on the WAR before making it available for download? If so, any >>>>> way to disable that (and, thus, free devs from having to wait for 3 min >>>>> while mvn deploy's running). >>>>> >>>>> For comparison, scp'ing the same file between from the same dev machine to >>>>> the Nexus box took < 19s. >>>>> >>>>> thanks! >>>>> >>>>> -nikita >>>>> >>>>> ----- >>>>> ---------------- >>>>> Nikita Tovstoles >>>>> vside.com >>>>> ---------------- >>>>> >>>>> -- >>>>> View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >>>>> Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. >>>>> >>>>> >>>>> --------------------------------------------------------------------- >>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>> For additional commands, e-mail: nexus-user-help@... >>>>> >>>>> >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>> For additional commands, e-mail: nexus-user-help@... >>>> >>>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>> For additional commands, e-mail: nexus-user-help@... >>> >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?@Brian: We're using 2.0.9 as well and though I know that 2.2.1 exists, it's no trivial matter for us to upgrade at the moment. re: speed - when 'deploy' phase runs, mvn displays upload progress as X/Y (x byte of y uploaded) and it reaches 100% in < 5. There is no proxy. Also, if it were a networking issue, why does scp complete in < 19 seconds. I will try bouncing Nexus and report back whether times improve but would appreciate any other suggestions.
----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?A tip: monitor the maven's JVM for memory, since as far as I remember, the java.net.URL (used by lightweight wagon) does some weirdo things with caching-in-memory or sort of...
~t~
On Tue, Nov 10, 2009 at 5:09 PM, dukehoops <nikita.tovstoles@...> wrote:
|
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?On Tue, Nov 10, 2009 at 11:09 AM, dukehoops <nikita.tovstoles@...> wrote:
> > > @Brian: > > We're using 2.0.9 as well and though I know that 2.2.1 exists, it's no > trivial matter for us to upgrade at the moment. > > re: speed - when 'deploy' phase runs, mvn displays upload progress as X/Y (x > byte of y uploaded) and it reaches 100% in < 5. There is no proxy. Also, if > it were a networking issue, why does scp complete in < 19 seconds. > > I will try bouncing Nexus and report back whether times improve but would > appreciate any other suggestions. > > Are you on a vm where disk io might be an issue? > Stephen Connolly-2 wrote: >> >> 2.0.9, 2.0.10, 2.2.1 >> >> 2009/11/10 Brian Fox <brianf@...>: >>> What version of Maven? >>> >>> On Tue, Nov 10, 2009 at 3:47 AM, Stephen Connolly >>> <stephen.alan.connolly@...> wrote: >>>> Well we were getting the upload of our 100MB+ zip file in < 5 seconds >>>> and then the mvn prompt sitting idle for 3-4 minutes before continuing >>>> w/o error. It was only when we went to a bigger zip file that we got >>>> the error. >>>> >>>> So that sounds like the same thing! ;-) >>>> >>>> -Stephen >>>> >>>> 2009/11/10 Brian Fox <brianf@...>: >>>>> These don't sound related. Specifically in Nexus 1.3.6+, the files are >>>>> first written to the .tmp location as they stream in, then it's >>>>> renamed to the final location. If you are seeing it slowly come in, >>>>> this tells me it's not the indexing, but something between you and >>>>> Nexus. Are you running Nexus on a VM or behind some reverse proxy? >>>>> >>>>> On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly >>>>> <stephen.alan.connolly@...> wrote: >>>>>> Sounds exactly like the behaviour we had with >>>>>> https://issues.sonatype.org/browse/NEXUS-2879 which we have >>>>>> encountered when our nexus has been running for a while. Restarting >>>>>> nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we >>>>>> get a chance we are planning to upgrade. Until then the issue has not >>>>>> repeated for us (I think it requires a long uptime with lots of big >>>>>> deployments >>>>>> >>>>>> -Stephen >>>>>> >>>>>> 2009/11/9 dukehoops <nikita.tovstoles@...>: >>>>>>> >>>>>>> Hi, >>>>>>> >>>>>>> We're running Nexus 1.3.6 on a separate linux box on our intranet. >>>>>>> During >>>>>>> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn >>>>>>> prompt >>>>>>> sits idle for 3 minutes, eventually carrying on w/o an error. During >>>>>>> those >>>>>>> 3-4 minutes: >>>>>>> - viewing that repo in a browser does NOT list the new WAR file >>>>>>> - 'ls -la' on hosted repo's directory shows something slowly writing >>>>>>> out >>>>>>> myartifact.war.tmp. Shortly after the .tmp file is fully written (== >>>>>>> in size >>>>>>> to the war), the actual WAR is created and web listing of the repo is >>>>>>> updated. >>>>>>> >>>>>>> The repo's configured with "Include in Search = false" - so I am >>>>>>> assuming >>>>>>> indexing should not be happening. Does Nexus do some sort of other >>>>>>> processing on the WAR before making it available for download? If so, >>>>>>> any >>>>>>> way to disable that (and, thus, free devs from having to wait for 3 >>>>>>> min >>>>>>> while mvn deploy's running). >>>>>>> >>>>>>> For comparison, scp'ing the same file between from the same dev >>>>>>> machine to >>>>>>> the Nexus box took < 19s. >>>>>>> >>>>>>> thanks! >>>>>>> >>>>>>> -nikita >>>>>>> >>>>>>> ----- >>>>>>> ---------------- >>>>>>> Nikita Tovstoles >>>>>>> vside.com >>>>>>> ---------------- >>>>>>> >>>>>>> -- >>>>>>> View this message in context: >>>>>>> http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >>>>>>> Sent from the Nexus Maven Repository Manager Users List mailing list >>>>>>> archive at Nabble.com. >>>>>>> >>>>>>> >>>>>>> --------------------------------------------------------------------- >>>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>>> >>>>>>> >>>>>> >>>>>> --------------------------------------------------------------------- >>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>> >>>>>> >>>>> >>>>> --------------------------------------------------------------------- >>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>> For additional commands, e-mail: nexus-user-help@... >>>>> >>>>> >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>> For additional commands, e-mail: nexus-user-help@... >>>> >>>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>> For additional commands, e-mail: nexus-user-help@... >>> >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> >> > > > ----- > ---------------- > Nikita Tovstoles > vside.com > ---------------- > > -- > View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26286009.html > Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Are you saying mvn's upload progress indicator may not reflect actual network traffic (by 60x)?
----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?I would also try to watch your network traffic on the machine doing
the deploy, I will have to check the maven code but I think Tamas is right, the upload indicator in maven is misleading. It is reading and buffering the data into memory so Maven thinks it's been sent, but in the meantime it's more slowly going over the wire. I just double checked with Tamas that if you see the .tmp file being slowly written, this is a key indicator. Nexus is getting the stream from Jetty and writing it directly to disk. Only after that point do we begin to process the checksums and indexes, so it's unlikely that processing is causing the slowdown. The reason I asked about Maven is because in 2.1.0 we tried switching to the http-commons wagon and found that it wasn't handling pre-emptive auth correctly and was actually uploading the data twice. On Tue, Nov 10, 2009 at 11:19 AM, Brian Fox <brianf@...> wrote: > On Tue, Nov 10, 2009 at 11:09 AM, dukehoops <nikita.tovstoles@...> wrote: >> >> >> @Brian: >> >> We're using 2.0.9 as well and though I know that 2.2.1 exists, it's no >> trivial matter for us to upgrade at the moment. >> >> re: speed - when 'deploy' phase runs, mvn displays upload progress as X/Y (x >> byte of y uploaded) and it reaches 100% in < 5. There is no proxy. Also, if >> it were a networking issue, why does scp complete in < 19 seconds. >> >> I will try bouncing Nexus and report back whether times improve but would >> appreciate any other suggestions. >> >> > > > Are you on a vm where disk io might be an issue? > >> Stephen Connolly-2 wrote: >>> >>> 2.0.9, 2.0.10, 2.2.1 >>> >>> 2009/11/10 Brian Fox <brianf@...>: >>>> What version of Maven? >>>> >>>> On Tue, Nov 10, 2009 at 3:47 AM, Stephen Connolly >>>> <stephen.alan.connolly@...> wrote: >>>>> Well we were getting the upload of our 100MB+ zip file in < 5 seconds >>>>> and then the mvn prompt sitting idle for 3-4 minutes before continuing >>>>> w/o error. It was only when we went to a bigger zip file that we got >>>>> the error. >>>>> >>>>> So that sounds like the same thing! ;-) >>>>> >>>>> -Stephen >>>>> >>>>> 2009/11/10 Brian Fox <brianf@...>: >>>>>> These don't sound related. Specifically in Nexus 1.3.6+, the files are >>>>>> first written to the .tmp location as they stream in, then it's >>>>>> renamed to the final location. If you are seeing it slowly come in, >>>>>> this tells me it's not the indexing, but something between you and >>>>>> Nexus. Are you running Nexus on a VM or behind some reverse proxy? >>>>>> >>>>>> On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly >>>>>> <stephen.alan.connolly@...> wrote: >>>>>>> Sounds exactly like the behaviour we had with >>>>>>> https://issues.sonatype.org/browse/NEXUS-2879 which we have >>>>>>> encountered when our nexus has been running for a while. Restarting >>>>>>> nexus fixed the issue. It's claimed to be fixed in 1.4.0, so when we >>>>>>> get a chance we are planning to upgrade. Until then the issue has not >>>>>>> repeated for us (I think it requires a long uptime with lots of big >>>>>>> deployments >>>>>>> >>>>>>> -Stephen >>>>>>> >>>>>>> 2009/11/9 dukehoops <nikita.tovstoles@...>: >>>>>>>> >>>>>>>> Hi, >>>>>>>> >>>>>>>> We're running Nexus 1.3.6 on a separate linux box on our intranet. >>>>>>>> During >>>>>>>> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then the mvn >>>>>>>> prompt >>>>>>>> sits idle for 3 minutes, eventually carrying on w/o an error. During >>>>>>>> those >>>>>>>> 3-4 minutes: >>>>>>>> - viewing that repo in a browser does NOT list the new WAR file >>>>>>>> - 'ls -la' on hosted repo's directory shows something slowly writing >>>>>>>> out >>>>>>>> myartifact.war.tmp. Shortly after the .tmp file is fully written (== >>>>>>>> in size >>>>>>>> to the war), the actual WAR is created and web listing of the repo is >>>>>>>> updated. >>>>>>>> >>>>>>>> The repo's configured with "Include in Search = false" - so I am >>>>>>>> assuming >>>>>>>> indexing should not be happening. Does Nexus do some sort of other >>>>>>>> processing on the WAR before making it available for download? If so, >>>>>>>> any >>>>>>>> way to disable that (and, thus, free devs from having to wait for 3 >>>>>>>> min >>>>>>>> while mvn deploy's running). >>>>>>>> >>>>>>>> For comparison, scp'ing the same file between from the same dev >>>>>>>> machine to >>>>>>>> the Nexus box took < 19s. >>>>>>>> >>>>>>>> thanks! >>>>>>>> >>>>>>>> -nikita >>>>>>>> >>>>>>>> ----- >>>>>>>> ---------------- >>>>>>>> Nikita Tovstoles >>>>>>>> vside.com >>>>>>>> ---------------- >>>>>>>> >>>>>>>> -- >>>>>>>> View this message in context: >>>>>>>> http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >>>>>>>> Sent from the Nexus Maven Repository Manager Users List mailing list >>>>>>>> archive at Nabble.com. >>>>>>>> >>>>>>>> >>>>>>>> --------------------------------------------------------------------- >>>>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>>>> >>>>>>>> >>>>>>> >>>>>>> --------------------------------------------------------------------- >>>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>>> >>>>>>> >>>>>> >>>>>> --------------------------------------------------------------------- >>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>> >>>>>> >>>>> >>>>> --------------------------------------------------------------------- >>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>> For additional commands, e-mail: nexus-user-help@... >>>>> >>>>> >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>> For additional commands, e-mail: nexus-user-help@... >>>> >>>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>> For additional commands, e-mail: nexus-user-help@... >>> >>> >>> >> >> >> ----- >> ---------------- >> Nikita Tovstoles >> vside.com >> ---------------- >> >> -- >> View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26286009.html >> Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. >> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?On Tue, Nov 10, 2009 at 11:24 AM, dukehoops <nikita.tovstoles@...> wrote:
> > Are you saying mvn's upload progress indicator may not reflect actual network > traffic (by 60x)? Yes, that's my suspicion. What OS is your server and client on? We have seen that some tweaks to the jetty config can improve performance if the server is on Windows. --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Doubt it - though not sure how to check. How can I check that? Also, wouldn't a disk IO constraint also apply to plain scp? we are using a funky vm (on dev box side):
D:\Views\main\v3classifieds\c4war>java -version java version "1.5.0" Java(TM) 2 Runtime Environment, Standard Edition (build pwi32dev-20080710a (SR8)) IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Windows Server 2003 x86-32 j9vmwi3223-20080710 (JIT enabled) J9VM - 20080625_20583_lHdSMr JIT - 20080620_1845_r8 GC - 200806_19) JCL - 20080710
----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?For kicks you could try the sun JVM
On Nov 10, 2009, at 11:29 AM, dukehoops wrote: > > Doubt it - though not sure how to check. How can I check that? Also, > wouldn't > a disk IO constraint also apply to plain scp? we are using a funky > vm (on > dev box side): > > D:\Views\main\v3classifieds\c4war>java -version > java version "1.5.0" > Java(TM) 2 Runtime Environment, Standard Edition (build > pwi32dev-20080710a > (SR8)) > IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Windows Server 2003 x86-32 > j9vmwi3223-20080710 (JIT enabled) > J9VM - 20080625_20583_lHdSMr > JIT - 20080620_1845_r8 > GC - 200806_19) > JCL - 20080710 > > > Brian Fox wrote: >> >> On Tue, Nov 10, 2009 at 11:09 AM, dukehoops <nikita.tovstoles@... >> > >> wrote: >>> >>> >>> @Brian: >>> >>> We're using 2.0.9 as well and though I know that 2.2.1 exists, >>> it's no >>> trivial matter for us to upgrade at the moment. >>> >>> re: speed - when 'deploy' phase runs, mvn displays upload progress >>> as X/Y >>> (x >>> byte of y uploaded) and it reaches 100% in < 5. There is no proxy. >>> Also, >>> if >>> it were a networking issue, why does scp complete in < 19 seconds. >>> >>> I will try bouncing Nexus and report back whether times improve >>> but would >>> appreciate any other suggestions. >>> >>> >> >> >> Are you on a vm where disk io might be an issue? >> >>> Stephen Connolly-2 wrote: >>>> >>>> 2.0.9, 2.0.10, 2.2.1 >>>> >>>> 2009/11/10 Brian Fox <brianf@...>: >>>>> What version of Maven? >>>>> >>>>> On Tue, Nov 10, 2009 at 3:47 AM, Stephen Connolly >>>>> <stephen.alan.connolly@...> wrote: >>>>>> Well we were getting the upload of our 100MB+ zip file in < 5 >>>>>> seconds >>>>>> and then the mvn prompt sitting idle for 3-4 minutes before >>>>>> continuing >>>>>> w/o error. It was only when we went to a bigger zip file that >>>>>> we got >>>>>> the error. >>>>>> >>>>>> So that sounds like the same thing! ;-) >>>>>> >>>>>> -Stephen >>>>>> >>>>>> 2009/11/10 Brian Fox <brianf@...>: >>>>>>> These don't sound related. Specifically in Nexus 1.3.6+, the >>>>>>> files >>>>>>> are >>>>>>> first written to the .tmp location as they stream in, then it's >>>>>>> renamed to the final location. If you are seeing it slowly >>>>>>> come in, >>>>>>> this tells me it's not the indexing, but something between you >>>>>>> and >>>>>>> Nexus. Are you running Nexus on a VM or behind some reverse >>>>>>> proxy? >>>>>>> >>>>>>> On Mon, Nov 9, 2009 at 3:13 PM, Stephen Connolly >>>>>>> <stephen.alan.connolly@...> wrote: >>>>>>>> Sounds exactly like the behaviour we had with >>>>>>>> https://issues.sonatype.org/browse/NEXUS-2879 which we have >>>>>>>> encountered when our nexus has been running for a while. >>>>>>>> Restarting >>>>>>>> nexus fixed the issue. It's claimed to be fixed in 1.4.0, so >>>>>>>> when >>>>>>>> we >>>>>>>> get a chance we are planning to upgrade. Until then the issue >>>>>>>> has >>>>>>>> not >>>>>>>> repeated for us (I think it requires a long uptime with lots >>>>>>>> of big >>>>>>>> deployments >>>>>>>> >>>>>>>> -Stephen >>>>>>>> >>>>>>>> 2009/11/9 dukehoops <nikita.tovstoles@...>: >>>>>>>>> >>>>>>>>> Hi, >>>>>>>>> >>>>>>>>> We're running Nexus 1.3.6 on a separate linux box on our >>>>>>>>> intranet. >>>>>>>>> During >>>>>>>>> 'mvn deploy' a 40m WAR upload take s < 5 seconds but then >>>>>>>>> the mvn >>>>>>>>> prompt >>>>>>>>> sits idle for 3 minutes, eventually carrying on w/o an error. >>>>>>>>> During >>>>>>>>> those >>>>>>>>> 3-4 minutes: >>>>>>>>> - viewing that repo in a browser does NOT list the new WAR >>>>>>>>> file >>>>>>>>> - 'ls -la' on hosted repo's directory shows something slowly >>>>>>>>> writing >>>>>>>>> out >>>>>>>>> myartifact.war.tmp. Shortly after the .tmp file is fully >>>>>>>>> written >>>>>>>>> (== >>>>>>>>> in size >>>>>>>>> to the war), the actual WAR is created and web listing of >>>>>>>>> the repo >>>>>>>>> is >>>>>>>>> updated. >>>>>>>>> >>>>>>>>> The repo's configured with "Include in Search = false" - so >>>>>>>>> I am >>>>>>>>> assuming >>>>>>>>> indexing should not be happening. Does Nexus do some sort of >>>>>>>>> other >>>>>>>>> processing on the WAR before making it available for >>>>>>>>> download? If >>>>>>>>> so, >>>>>>>>> any >>>>>>>>> way to disable that (and, thus, free devs from having to >>>>>>>>> wait for 3 >>>>>>>>> min >>>>>>>>> while mvn deploy's running). >>>>>>>>> >>>>>>>>> For comparison, scp'ing the same file between from the same >>>>>>>>> dev >>>>>>>>> machine to >>>>>>>>> the Nexus box took < 19s. >>>>>>>>> >>>>>>>>> thanks! >>>>>>>>> >>>>>>>>> -nikita >>>>>>>>> >>>>>>>>> ----- >>>>>>>>> ---------------- >>>>>>>>> Nikita Tovstoles >>>>>>>>> vside.com >>>>>>>>> ---------------- >>>>>>>>> >>>>>>>>> -- >>>>>>>>> View this message in context: >>>>>>>>> http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26274731.html >>>>>>>>> Sent from the Nexus Maven Repository Manager Users List >>>>>>>>> mailing >>>>>>>>> list >>>>>>>>> archive at Nabble.com. >>>>>>>>> >>>>>>>>> >>>>>>>>> --------------------------------------------------------------------- >>>>>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>>> --------------------------------------------------------------------- >>>>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>>>> >>>>>>>> >>>>>>> >>>>>>> --------------------------------------------------------------------- >>>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>>> >>>>>>> >>>>>> >>>>>> --------------------------------------------------------------------- >>>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>>> For additional commands, e-mail: nexus-user-help@... >>>>>> >>>>>> >>>>> >>>>> --------------------------------------------------------------------- >>>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>>> For additional commands, e-mail: nexus-user-help@... >>>>> >>>>> >>>> >>>> --------------------------------------------------------------------- >>>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>>> For additional commands, e-mail: nexus-user-help@... >>>> >>>> >>>> >>> >>> >>> ----- >>> ---------------- >>> Nikita Tovstoles >>> vside.com >>> ---------------- >>> >>> -- >>> View this message in context: >>> http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26286009.html >>> Sent from the Nexus Maven Repository Manager Users List mailing list >>> archive at Nabble.com. >>> >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: nexus-user-unsubscribe@... >>> For additional commands, e-mail: nexus-user-help@... >>> >>> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> >> > > > ----- > ---------------- > Nikita Tovstoles > vside.com > ---------------- > > -- > View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26286553.html > Sent from the Nexus Maven Repository Manager Users List mailing list > archive at Nabble.com. > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Client:
win server 2003 x64 SP2; 3g Xeon, 8gb ram, SCSI D:\Views\main\v3classifieds\c4war>java -version java version "1.5.0" Java(TM) 2 Runtime Environment, Standard Edition (build pwi32dev-20080710a (SR8)) IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Windows Server 2003 x86-32 j9vmwi3223-20080710 (JIT enabled) J9VM - 20080625_20583_lHdSMr JIT - 20080620_1845_r8 GC - 200806_19) JCL - 20080710 SERVER: 4 2xCore Xeon, RHEL 2.6.9-78.0.8.ELsmp, not sure what kind of disk (but it's big ;-) Here's how nexus's running: [root@sjd-pdvas-005 ~]# ps aux | grep nexus nexus 11932 0.0 0.0 11916 620 ? Sl Aug26 0:00 /opt/nexus/nexus-webapp-1.3.6/bin/jsw/linux-x86-32/./wrapper /opt/nexus/nexus-webapp-1.3.6/bin/jsw/linux-x86-32/../../../conf/wrapper.conf wrapper.syslog.ident=nexus-webapp wrapper.pidfile=/opt/nexus/nexus-webapp-1.3.6/bin/jsw/linux-x86-32/./nexus-webapp.pid wrapper.daemonize=TRUE nexus 11944 0.0 2.1 612628 175648 ? Sl Aug26 19:05 /opt/c4/java/bin/java -Dbasedir=../../.. -Djava.io.tmpdir=../../../runtime/tmp -Dsun.net.inetaddr.ttl=3600 -Djava.library.path=./lib -classpath ../../../lib/plexus-platform-jsw-1.6.6.jar:../../../conf/ -Dwrapper.key=QpVAMS4usm08Tv4G -Dwrapper.port=32000 -Dwrapper.jvm.port.min=31000 -Dwrapper.jvm.port.max=31999 -Dwrapper.pid=11932 -Dwrapper.version=3.2.3 -Dwrapper.native_library=wrapper -Dwrapper.service=TRUE -Dwrapper.cpu.timeout=10 -Dwrapper.jvmid=1 org.codehaus.plexus.classworlds.launcher.Launcher ...using this VM: java version "1.5.0" Java(TM) 2 Runtime Environment, Standard Edition (build pxi32dev-20080315 (SR7)) IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Linux x86-32 j9vmxi3223-20080315 (JIT enabled) J9VM - 20080314_17962_lHdSMr JIT - 20080130_0718ifx2_r8 GC - 200802_08) JCL - 20080314
----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?FWIW, I've seen a delay with large artifacts as well.
In my observation the files are actually done uploading when the progress indicator reaches it's maximum, if you look in the Nexus storage directory you will see them. What seems to be happening is that the checksum calculation (md5 and sha1) takes a long time for large files, and Maven doesn't proceed until this has finished. Perhaps someone can confirm? Rich On Nov 10, 2009, at 8:37 AM, dukehoops wrote: > > Client: > > win server 2003 x64 SP2; 3g Xeon, 8gb ram, SCSI > > D:\Views\main\v3classifieds\c4war>java -version > java version "1.5.0" > Java(TM) 2 Runtime Environment, Standard Edition (build pwi32dev-20080710a > (SR8)) > IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Windows Server 2003 x86-32 > j9vmwi3223-20080710 (JIT enabled) > J9VM - 20080625_20583_lHdSMr > JIT - 20080620_1845_r8 > GC - 200806_19) > JCL - 20080710 > > SERVER: > 4 2xCore Xeon, RHEL 2.6.9-78.0.8.ELsmp, not sure what kind of disk (but it's > big ;-) Here's how nexus's running: > > [root@sjd-pdvas-005 ~]# ps aux | grep nexus > nexus 11932 0.0 0.0 11916 620 ? Sl Aug26 0:00 > /opt/nexus/nexus-webapp-1.3.6/bin/jsw/linux-x86-32/./wrapper > /opt/nexus/nexus-webapp-1.3.6/bin/jsw/linux-x86-32/../../../conf/wrapper.conf > wrapper.syslog.ident=nexus-webapp > wrapper.pidfile=/opt/nexus/nexus-webapp-1.3.6/bin/jsw/linux-x86-32/./nexus-webapp.pid > wrapper.daemonize=TRUE > nexus 11944 0.0 2.1 612628 175648 ? Sl Aug26 19:05 > /opt/c4/java/bin/java -Dbasedir=../../.. > -Djava.io.tmpdir=../../../runtime/tmp -Dsun.net.inetaddr.ttl=3600 > -Djava.library.path=./lib -classpath > ../../../lib/plexus-platform-jsw-1.6.6.jar:../../../conf/ > -Dwrapper.key=QpVAMS4usm08Tv4G -Dwrapper.port=32000 > -Dwrapper.jvm.port.min=31000 -Dwrapper.jvm.port.max=31999 > -Dwrapper.pid=11932 -Dwrapper.version=3.2.3 -Dwrapper.native_library=wrapper > -Dwrapper.service=TRUE -Dwrapper.cpu.timeout=10 -Dwrapper.jvmid=1 > org.codehaus.plexus.classworlds.launcher.Launcher > > ...using this VM: > > java version "1.5.0" > Java(TM) 2 Runtime Environment, Standard Edition (build pxi32dev-20080315 > (SR7)) > IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Linux x86-32 j9vmxi3223-20080315 > (JIT enabled) > J9VM - 20080314_17962_lHdSMr > JIT - 20080130_0718ifx2_r8 > GC - 200802_08) > JCL - 20080314 > > > > Brian Fox wrote: >> >> On Tue, Nov 10, 2009 at 11:24 AM, dukehoops <nikita.tovstoles@...> >> wrote: >>> >>> Are you saying mvn's upload progress indicator may not reflect actual >>> network >>> traffic (by 60x)? >> >> Yes, that's my suspicion. What OS is your server and client on? We >> have seen that some tweaks to the jetty config can improve performance >> if the server is on Windows. >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: nexus-user-unsubscribe@... >> For additional commands, e-mail: nexus-user-help@... >> >> >> > > > ----- > ---------------- > Nikita Tovstoles > vside.com > ---------------- > > -- > View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26286648.html > Sent from the Nexus Maven Repository Manager Users List mailing list archive at Nabble.com. > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: nexus-user-unsubscribe@... > For additional commands, e-mail: nexus-user-help@... > --------------------------------------------------------------------- To unsubscribe, e-mail: nexus-user-unsubscribe@... For additional commands, e-mail: nexus-user-help@... |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?I was hoping to avoid "kicks" ;- So, to summarize:
Hypothesis A: the upload is actually taking much longer than mvn deploy's progress indicator shows (check by watching the network pipe) Hypothesis B: File gets uploaded to server Jetty's heap space quickly, but Jetty (or whatever's running on Jetty) is taking its' sweet time writing the file out to disk. I noticed that our config does not explicitly set GC policy and heap sizes. So, we're relying on J9's ergonomics and who knows what those are. Any recommendations for GC policy and heap sizing along the lines of: http://confluence.atlassian.com/display/DOC/Garbage+Collector+Performance+Issues ? I'd imagine Nexus webapp would benefit from relatively large eden (new gen) space? thanks -nikita
----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?Not sure I can confirm anything - I am a Nexus newbie - but here's a Nexus debug file covering a single mvn deploy-triggered upload of a 42mb WAR artifact (same as described in my initial post). Things to note:
-upload kicks off at 10:02:43 -on the client side, mvn deploy's upload indicator reaches 100% about 3 seconds later -on server side EOF takes place at 10:03:39 - (about a minute later) -- right above, the last two lines of 02:43 complain about 'perms/roles' but the upload seems to proceed successfully anyway (What's that about? the user in question has 'deployment' role for all repos) - then, from 03:39 to 04:37 (another minute) something's copying the war from somewhere to repo storage dir. Where is "/com/kijiji/c4war/4.0.e000-ntovstoles/c4war-4.0.e000-ntovstoles.war" relative to nexus home? - we're done at 04:39 Log attached: nexus-debug.log
----------------
Nikita Tovstoles vside.com ---------------- |
|
|
Re: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?@03:39: the WAR upload begins (Jetty got REQUEST) -- "from somewhere" is the HTTP Request body of your Maven (jetty log states that it just got a request)
@04:37: the WAR upload ends (Jetty sends back RESPONSE 201 "created") -- file should be written to disk completely That's 58 seconds for the WAR. The other "gross" time you see is all the other "maven nitty-bitty" stuff, like sha1+md5, metadata and POM deployment. At the log beginning, there is some request ending with a 401 (probably Maven tries to act without authenticating), then there is a pause of appx a minute (02:43 - 03:39) where Nexus not having any incoming request. Probably Maven is working locally.
Hope helps, ~t~ On Tue, Nov 10, 2009 at 6:18 PM, dukehoops <nikita.tovstoles@...> wrote:
|
|
|
RE: mvn deploy to hosted repo: 5 sec to upload, 3 min to process?We see the same issues when uploading 50mb war files. We
generally see total times of 3 - 4 minutes. In our case, the entire time is
spent writing to the tmp war file. The move / rename to the actual war file
seems instantaneous and maven cli becomes responsive immediately afterwards. An
ftp of the same file among the same servers take 15 seconds. From:
t.cservenak@... [mailto:t.cservenak@...] On Behalf Of Tamás
Cservenák @03:39: the WAR upload begins (Jetty got REQUEST) --
"from somewhere" is the HTTP Request body of your Maven (jetty log
states that it just got a request) @04:37: the WAR upload ends (Jetty
sends back RESPONSE 201 "created") -- file should be written to disk
completely That's 58 seconds for the WAR. The other "gross"
time you see is all the other "maven nitty-bitty" stuff, like
sha1+md5, metadata and POM deployment. At the log beginning, there is some request ending with a
401 (probably Maven tries to act without authenticating), then there is a pause
of appx a minute (02:43 - 03:39) where Nexus not having any incoming request.
Probably Maven is working locally. Hope helps, ~t~ On Tue, Nov 10, 2009 at 6:18 PM, dukehoops <nikita.tovstoles@...>
wrote:
View this message in context: http://old.nabble.com/mvn-deploy-to-hosted-repo%3A-5-sec-to-upload%2C-3-min-to-process--tp26274731p26287334.html Sent from the Nexus Maven
Repository Manager Users List mailing list archive at Nabble.com. |
| < Prev | 1 - 2 | Next > |
| Free embeddable forum powered by Nabble | Forum Help |