Nabble - DRBD - User

Concurrent writes

2009-12-18T12:14:45Z

Hello list,
i have found an old tread about our problem back from April and the last
message from October (at the end of the email, so sorry for top posting) with
a hint to our setup.

The setup:
cluster with 2 nodes and 2 storages - Stor1 and Stor2 (part of the cluster =
4 nodes total)
several LVM volumes replicated via DRBD (8.3.2) in active/active and exported
via IET as blockio then imported from all cluster members (including the
storages) from both storages at the same time and multipath used in multibus
mode from all.

The problem:
Not tested with other DRBD versions, but at least with different kernels the
results are the same.
When cman is stopped on one/any of the storages i get in the logs for both of
them:
"block drbd0: istiod1[2739] Concurrent local write detected! [DISCARD L] new:
0s +1024; pending: 0s +1024"
it is always for drbd0 (actively used from all active nodes) and always for
0s +1024

My guess is that this is caused from multipath checking for the disk being
alive, but if so it should also happen when cman is running or for other
volumes too not?

P.S. please CC me as i am not subscribed to the list

------------ the last message in the tread --------------------------
Gennadiy,

I realize it's been a while since this issue was discussed last, but
we're still trying to hunt this down. I realize this is asking a lot,
but do you happen to still have the drbd.conf available from back in
April? It was unfortunately never posted in the thread.

Specifically, I'm curious as to whether you were using dual-Primary mode
at the time (allow-two-primaries).

Thanks!

Cheers,
Florian

On 2009-04-22 00:09, Gennadiy Nerubayev wrote:

> On Tue, Apr 21, 2009 at 5:39 PM, Roof, Morey R. <MRoof at admin.nmt.edu
> <mailto:MRoof at admin.nmt.edu>> wrote:
>
> The Initiators I use are: VMWare ESX 3.0.x, VMWare ESX 3.5.x, Micrsoft
> iSCSI Initiator 2.08, and Linux (RHEL 5, SUSE 10).
>
> I gave SQLIO a run with your params and I'm not getting the concurrent
> write issue.
>
>
> The bizarreness continues; downgrading DRBD is of no help either :(
> Although, it is encouraging in a way suggesting that this might be
> finally isolated. I assume you're checking the logs of the primary node?
>
> The hardware I run DRBD is a pair of HP Proliant DL380 G4 servers with
> 12GB of RAM and P600 SAS controllers. The machines are currently
> replicating 1.8TB of data. I run DRBD under SuSE 10SP2 and use IET
> stock from SuSE but the DRBD is 8.0.16.
>
>
> What is the kernel version on the target? Version of Windows on the
> initiator? Could you share the relevant portions of ietd.conf and drbd.conf?
>
> Thanks,
>
> -Gennadiy

_______________________________________________
drbd-dev mailing list
drbd-dev@...
http://lists.linbit.com/mailman/listinfo/drbd-dev

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

[Drbd-dev] Concurrent writes

2009-12-18T04:04:17Z

_______________________________________________
drbd-dev mailing list
drbd-dev@...
http://lists.linbit.com/mailman/listinfo/drbd-dev

Re: block drbd0: bio would need to, but cannot, be split error on physical device

2009-12-17T12:46:30Z

Thanks for the advice. It seems CONFIG_PPC_4K_PAGES is obsolete, because make oldconfig was removing it. What I did instead I set CONFIG_PPC_64K_PAGES=n and recompiled the kernel. I had to disable one patch and use --without kabichk option (for those who are in the same boat) to be able to create standard kernel RPM

Patch23256: linux-2.6-ppc64-subpage-protection-for-pave.patch

drbd now reports:

block drbd0: max_segment_size ( = BIO size ) = 32768

and I was able to create the file system.

Thanks again.

Sincerely yours,
Vadym Chepkov

--- On Thu, 12/17/09, Lars Ellenberg <lars.ellenberg@...> wrote:

> From: Lars Ellenberg <lars.ellenberg@...>
> Subject: Re: [DRBD-user] block drbd0: bio would need to, but cannot, be split error on physical device
> To: drbd-user@...
> Date: Thursday, December 17, 2009, 10:46 AM
> On Thu, Dec 17, 2009 at 04:47:14AM
> -0800, Vadym Chepkov wrote:
> > Hi,
> >
> > Sorry for being impatient. Is there any chance I can
> make DRBD to work
> > on PPC platform or is it "terminal"? Thanks.
>
> Easy fix: use a kernel with "CONFIG_PPC_4K_PAGES" (or 16K,
> if you
> prefer). RHEL default for ppc64 seems to be 64k
> pages,
> and DRBD currently breaks with that setting.
>
> we are working to get a real fix for that problem,
> but we have a lot of other things to do as well.
>
> --
> : Lars Ellenberg
> : LINBIT | Your Way to High Availability
> : DRBD/HA support and consulting http://www.linbit.com
>
> DRBD® and LINBIT® are registered trademarks of LINBIT,
> Austria.
> __
> please don't Cc me, but send to
> list -- I'm subscribed
> _______________________________________________
> drbd-user mailing list
> drbd-user@...
> http://lists.linbit.com/mailman/listinfo/drbd-user
>

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: block drbd0: bio would need to, but cannot, be split error on physical device

2009-12-17T07:46:47Z

On Thu, Dec 17, 2009 at 04:47:14AM -0800, Vadym Chepkov wrote:
> Hi,
>
> Sorry for being impatient. Is there any chance I can make DRBD to work
> on PPC platform or is it "terminal"? Thanks.

Easy fix: use a kernel with "CONFIG_PPC_4K_PAGES" (or 16K, if you
prefer). RHEL default for ppc64 seems to be 64k pages,
and DRBD currently breaks with that setting.

we are working to get a real fix for that problem,
but we have a lot of other things to do as well.

--
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
__
please don't Cc me, but send to list -- I'm subscribed
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Interoperability of 8.0.16 and 8.3.6

2009-12-17T07:05:28Z

Hi,

DRBD 8.0 says proto:86 and 8.3.6 says proto:86-91, so I would expect
these two to work nicely together. And this seems to be the case, except
when bringing them up in primary/primary mode.

The problems seems to be that after both devices are in primary/primary
and connected state, the 8.0.16 node wants to go back to WFBitMapS and
remains there, but the 8.3.6 node refuses it due to: "block drbd0:
unexpected cstate (Connected) in receive_bitmap". So 8.0.16 remains
stuck in WFBitmapS while 8.3.6 thinks it's in Connected.

Is this something that should work and is maybe a bug (drbd or our
configuration), or is it not something realistic to expect?

thanks in advance,
iustin
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: block drbd0: bio would need to, but cannot, be split error on physical device

2009-12-17T04:47:14Z

Hi,

Sorry for being impatient. Is there any chance I can make DRBD to work on PPC platform or is it "terminal"? Thanks.

Sincerely yours,
Vadym Chepkov

--- On Thu, 12/10/09, Vadym Chepkov <chepkov@...> wrote:

> From: Vadym Chepkov <chepkov@...>
> Subject: Re: [DRBD-user] block drbd0: bio would need to, but cannot, be split error on physical device
> To: drbd-user@...
> Date: Thursday, December 10, 2009, 10:56 AM
> No, it doesn't work at all, as I
> said, I can't even mkfs
>
> Sincerely yours,
> Vadym Chepkov
>
>
> --- On Thu, 12/10/09, Lars Ellenberg <lars.ellenberg@...>
> wrote:
>
> > From: Lars Ellenberg <lars.ellenberg@...>
> > Subject: Re: [DRBD-user] block drbd0: bio would need
> to, but cannot, be split error on physical device
> > To: drbd-user@...
> > Date: Thursday, December 10, 2009, 10:54 AM
> > On Thu, Dec 10, 2009 at 07:40:24AM
> > -0800, Vadym Chepkov wrote:
> > >
> > > --- On Thu, 12/10/09, Lars Ellenberg <lars.ellenberg@...>
> > wrote:
> > > >
> > > > No, I was refering to We as in DRBD
> Developers,
> > > > and could do as in "are able to code".
> > > > But I'd rather code a real fix than
> something
> > that may
> > > > appear to work
> > > > most of the time, and then possibly break
> in
> > strange ways
> > > > sometime later
> > > > when used in different ways.
> > > >
> > > > We won't forget this issue, of course, but
> for
> > now, sorry,
> > > > recent DRBD
> > > > does not work properly on ppc64 with 64k
> page
> > size.
> > >
> > > Oh, so I can't use DRBD there at all?
> >
> > Well, if it "works for you" once you ignore this
> message,
> > I'm fine with that.
> >
> > --
> > : Lars Ellenberg
> > : LINBIT | Your Way to High Availability
> > : DRBD/HA support and consulting http://www.linbit.com
> >
> > DRBD® and LINBIT® are registered trademarks of
> LINBIT,
> > Austria.
> > __
> > please don't Cc me, but send to
> > list -- I'm subscribed
> > _______________________________________________
> > drbd-user mailing list
> > drbd-user@...
> > http://lists.linbit.com/mailman/listinfo/drbd-user
> >
>

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Status of 8.0 series?

2009-12-16T09:28:35Z

Hi all,

I see that was released back in March and that there have been a couple
of commits afterward on the 8.0 tree; but the last one was in late
April.

For someone who still needs to run 8.0, is it recommended to use 8.0.16
or the latest git?

thanks in advance,
iustin
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-16T08:40:35Z

Hi,

On 12/11/2009 11:19 AM, Mark Watts wrote:

On Thu, 2009-12-10 at 16:58 +0100, Florian Haas wrote:

Lars beat me to this but anyway...

On 2009-12-10 13:23, Beck Ingmar wrote:

Hi Florian,

why improves bonding of the replication-connection not the performance.

I did not say that. :) I said round-robin bonding _over 4 NICs_ would be
detrimental to performance.

We currently have a direct connection for the 1 Gb DRBD replication and
I
wanted to activate an active-active bonding (with a further LAN-port). I
would expect nearly double of replication-performance.

All of the following applies to DRBD Connected in protocol C:

- With 1 GBit Ethernet link, you can do about 110 MB/s (provided your
storage subsystem has that bandwidth);

- With 2 GBit links bonded in round-robin mode, you top out at around
160-170 MB/s (even then you may have to tune kernel TCP parameters);

- with 3 Gbit links, your throughput will drop down to around 110 MB/s;

- with 4 links or more, you will get _significantly less_ throughput
than a single unbonded link.

Reason for all this: TCP reordering.

Or depends it on whether the connection between the nodes directly or
via 1-n switches.

I am talking about back-to-back connections. Switches could make it worse.

Do Jumbo Frames make any difference here?

Yes they help, in both cases (with switch and in b2b scenario). Please test everything and put everything working without jumbo frames. Use jumbo frames only if you need extra performance and after you have everything working.

But there are some other settings, search in google for "linux kernel tcp optimizations".

Mark.


_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

-- 
Igor Neves igor.neves@...
3GNTW - Tecnologias de Informação, Lda
 
 SIP: igor@...	JID: igor@... 
 ICQ: 249075444		MSN: igor@...
 TLM: 00351914503611	PSTN: 00351252377120

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-16T08:36:42Z

Hi,

On 12/10/2009 10:05 PM, Jakov Sosic wrote:

> On Thu, 2009-12-10 at 19:11 +0100, Lars Ellenberg wrote:
>
>
>> exactly same hardware?
>>
> yes
>
>
>> exactly same kernel and stuff?
>>
> CentOS 5.4, I really look after details, and 'rpm -qa | sort' give me
> exactly the same output on both nodes, config files are also same,
> everything is OK.
>
> I tried reversing roles, and when secondary becomes primary, then the IO
> load reverses too.
>
> I found on the internet that 3ware controllers have issues on Linux, and
> I have 3ware controller... So maybe that is the part of problem...
>

Yes, they really suck, use areca raid controllers, and you will see what
it's a real hardware RAID.

(Yes, much better then software raid and 3ware controllers)

>
>
>> io scheduler? -> use deadline.
>>
> I'm using it... I know about kernel threads problem with CFQ.
>
>
>
>> otherwise: we sell "DRBD Healthchecks" ;)
>>
> I'm just a contractor too, but I think that the owners will buy FC
> storage and get on without cheap JBODs...
>
>
>
>

--
Igor Neves<igor.neves@...>
3GNTW - Tecnologias de Informação, Lda

SIP: igor@... JID: igor@...
ICQ: 249075444 MSN: igor@...
TLM: 00351914503611 PSTN: 00351252377120

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: Request about gis

2009-12-16T05:17:38Z

Am Mittwoch, 16. Dezember 2009 13:32:35 schrieb Michael Schwartzkopff:

> Hi,
>
> I read about the gis in http://www.drbd.org/docs/more and I did not find
> exactly the answer to my problem:
>
> I have to nodes and somebody (not me!) pulled the connection betweeen both
> nodes. I have a resource r0. The gis of the both nodes A and B are:
>
> On A:
> 63678445E843CFAE:0000000000000000:8F1D3DACC1B146D6:653C6F9D3EA0C1E7:1:1:1:0
>:0:0
>
> On B:
> 529517DB7A6B1F01:63678445E843CFAE:8F1D3DACC1B146D6:653C6F9D3EA0C1E7:1:1:1:0
>:0:0
>
> In this case I expect a full sync from B to A if I connect both nodes
> again. Is this correct?
>
> The next resource has the following gis:
>
> On node A:
> 8CEE141BCEC19430:0000000000000000:BD100C642C53F3D0:97C5BEF0622ABACE:1:1:0:0
>:0:0
>
> On node B:
> 439629C26573E473:8CEE141BCEC19431:BD100C642C53F3D0:97C5BEF0622ABACE:1:1:1:0
>:0:0
>
> Please note that the bitmaps differ exactly by "1".
>
> Can anybody explain me what would happen here, if I connect both DRBDs
> again? Thanks.
>
> Greetings,

Answer from the IRC:

Case 1) Sync from node B to node A.
Case 2) Same. The difference only show which was primary before (here: B).

Thanks!

--
Dr. Michael Schwartzkopff
MultiNET Services GmbH
Addresse: Bretonischer Ring 7; 85630 Grasbrunn; Germany
Tel: +49 - 89 - 45 69 11 0
Fax: +49 - 89 - 45 69 11 21
mob: +49 - 174 - 343 28 75

mail: misch@...
web: www.multinet.de

Sitz der Gesellschaft: 85630 Grasbrunn
Registergericht: Amtsgericht München HRB 114375
Geschäftsführer: Günter Jurgeneit, Hubert Martens

---

PGP Fingerprint: F919 3919 FF12 ED5A 2801 DEA6 AA77 57A4 EDD8 979B
Skype: misch42
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Request about gis

2009-12-16T04:32:35Z

Hi,

I read about the gis in http://www.drbd.org/docs/more and I did not find
exactly the answer to my problem:

I have to nodes and somebody (not me!) pulled the connection betweeen both
nodes. I have a resource r0. The gis of the both nodes A and B are:

On A:
63678445E843CFAE:0000000000000000:8F1D3DACC1B146D6:653C6F9D3EA0C1E7:1:1:1:0:0:0

On B:
529517DB7A6B1F01:63678445E843CFAE:8F1D3DACC1B146D6:653C6F9D3EA0C1E7:1:1:1:0:0:0

In this case I expect a full sync from B to A if I connect both nodes again.
Is this correct?

The next resource has the following gis:

On node A:
8CEE141BCEC19430:0000000000000000:BD100C642C53F3D0:97C5BEF0622ABACE:1:1:0:0:0:0

On node B:
439629C26573E473:8CEE141BCEC19431:BD100C642C53F3D0:97C5BEF0622ABACE:1:1:1:0:0:0

Please note that the bitmaps differ exactly by "1".

Can anybody explain me what would happen here, if I connect both DRBDs again?
Thanks.

Greetings,

--
Dr. Michael Schwartzkopff
MultiNET Services GmbH
Addresse: Bretonischer Ring 7; 85630 Grasbrunn; Germany
Tel: +49 - 89 - 45 69 11 0
Fax: +49 - 89 - 45 69 11 21
mob: +49 - 174 - 343 28 75

mail: misch@...
web: www.multinet.de

Sitz der Gesellschaft: 85630 Grasbrunn
Registergericht: Amtsgericht München HRB 114375
Geschäftsführer: Günter Jurgeneit, Hubert Martens

---

PGP Fingerprint: F919 3919 FF12 ED5A 2801 DEA6 AA77 57A4 EDD8 979B
Skype: misch42
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Expansion DRBD configuration on Bound-replication

2009-12-16T02:55:51Z

Hello everyone!

I have a running cluster with DRBD replication over 1 crossover cable.
Now I want to do replication over a crossover-Bound (round-robin).

What do I have to make (after installing the NIC) so that neither my data on the DRBD devices go lost nor my configuration of the DRBD devices get unusable.

Regards

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-16T02:33:37Z

On 2009-12-16 03:48, Jakov Sosic wrote:

> Hello to all!
>
> I've found the bottlenecks, so let me start over.
>
> I used 3ware controllers and RAID10 over 4 SATA Barracudas. I've googled
> around and I've found that 3ware RAID performance really sucks for a
> controller of that type. I did have the write cache turned on, although
> I didn't have BBU.
>
> So first step first - I've switched from hardware raid 10 do Linux
> software raid 10. My performance increased noticeably, for almost 30% on
> all bonnie++ tests, and in one test even 400%. That was it - I was
> convinced. I've reconfigured both my drbd nodes to software RAID-10.

Neil, someone owes you a drink of your favorite beverage. :)

> Next, I've noticed that when drbd is disconnected, my performance
> doubles. So I've started to investigate why was that. I've found that
> problem was that drbd replication was going through the same 4
> round-robin bonded interfaces that iSCSI export was going. So, I've
> split that 4 interfaces to two bonds, both round-robin. One was for
> iSCSI, other for DRBD replication only. Now, I've got the same
> performance as when drbd was disconnected, with minor decrease (~5-10%).

That, most likely, has nothing to do with the fact that previously you
shared that same bonded link between DRBD and iSCSI, and you no longer
do. Instead, it's most probably due to your DRBD link now being bonded
over just 2 NICs, not 4. See my earlier posts in this thread about this
issue.

Florian

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

signature.asc (268 bytes) Download Attachment

Re: crossover or switch?

2009-12-15T22:16:51Z

On 16/12/2009 04:56, Jakov Sosic wrote:
> Hi.
>
> Is there any benefit of doing drbd replication through two crossover
> cables instead of through two switches? Can I except maybe lower latency
> or there is no benefit at all? Has anyone tested this?
>

Hiya

To lessen single points of failure, where I work, on our second nics, we
use cross over cables.

For my point, it works well.

HTH

Brent Clark
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

crossover or switch?

2009-12-15T18:56:47Z

Hi.

Is there any benefit of doing drbd replication through two crossover
cables instead of through two switches? Can I except maybe lower latency
or there is no benefit at all? Has anyone tested this?

--
| Jakov Sosic | ICQ: 28410271 | PGP: 0x965CAE2D |
=================================================================
| |

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-15T18:48:10Z

On Thu, 2009-12-10 at 19:11 +0100, Lars Ellenberg wrote:

> Then maybe you are simply looking in the wrong direction.
>
> exactly same hardware?
> exactly same kernel and stuff?
> io scheduler? -> use deadline.
>
> otherwise: we sell "DRBD Healthchecks" ;)

Hello to all!

I've found the bottlenecks, so let me start over.

I used 3ware controllers and RAID10 over 4 SATA Barracudas. I've googled
around and I've found that 3ware RAID performance really sucks for a
controller of that type. I did have the write cache turned on, although
I didn't have BBU.

So first step first - I've switched from hardware raid 10 do Linux
software raid 10. My performance increased noticeably, for almost 30% on
all bonnie++ tests, and in one test even 400%. That was it - I was
convinced. I've reconfigured both my drbd nodes to software RAID-10.

Next, I've noticed that when drbd is disconnected, my performance
doubles. So I've started to investigate why was that. I've found that
problem was that drbd replication was going through the same 4
round-robin bonded interfaces that iSCSI export was going. So, I've
split that 4 interfaces to two bonds, both round-robin. One was for
iSCSI, other for DRBD replication only. Now, I've got the same
performance as when drbd was disconnected, with minor decrease (~5-10%).

One thing that boders me left - it seems to me that I cannot utilize my
disks to 100% over the network, no matter what do I do with them or how
do I trash them. I would like to find out where's the next bottleneck,
although the results I get on Xen domU are now fantastic, 4x faster than
with hardware raid and drbd replication on the same bond as iSCSI. Also,
I had RAID6 on this crappy controller before, so this is like 40x
increase of performance :) But still, if I got this far, I want to go
even beyond... I would be really happy if I could thrash the disks to
the maximum, if that's possible. Although Xen over (C)LVM over iSCSI
over DRBD over mdraid is too much overlays, and any of them could cause
the slowdowns (although I suspect at iSCSI).

--
| Jakov Sosic | ICQ: 28410271 | PGP: 0x965CAE2D |
=================================================================
| |

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Steps to resize resources on top of LVM, quick peer review is needed.

2009-12-15T06:33:51Z

Hi everyone,

I need to resize an online drbd resource. Drbd is on top of LVM.

resource drbd0 {

on node1.aka.edu.au {
device /dev/drbd0;
disk /dev/lvmc1/export1;
address 192.168.1.1:7789;
meta-disk /dev/sdd1[0];
}

on node2.aka.edu.au {
device /dev/drbd0;
disk /dev/lvmc1/export1;
address 192.168.1.2:7789;
meta-disk /dev/sdd1[0];
}

Could someone who done this before confirm my steps below:

1. Add more space to Logical Group lvmc1 on node1 and node2

2. Extend logical volume /dev/lvmc1/export1 on node1 and node2 ( growing
from 1 TB to 1.5 TB )

3. Grow ext3 file systems using either

ext2online /dev/lvmc1/export1 or

reseizefs /dev/lvmc1/export1

4. drbdadm resize /|resource|/ ( in my case drbd0 ) on the primary node.

5. repeat for 1 - 4 for other resources.

Questions:

a. It is irrelevant on which node ( primary or secondary ) steps 1 & 2
are carried out first. Can you confirm this, pls?

b. Can you confirm that steps 3 and 4 are in correct order?

c. Can you confirm that both steps 3 and 4 can be done online?

d. Will step 4 cause any performance degradation? (How should it take?)

Thank you in advance.
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: Kernel Panic occuring when drbd is up & (re)syncing

2009-12-14T14:20:21Z

Hi list,

Any news about this bug ??

2009/11/16 Ivars Strazdiņš <ivars.strazdins@...>:

> It looks like I am getting kernel bug on 64-bit Xen Debian in similar
> conditions, ie, when running drbd-verify.
> I have got it happening on both cluster nodes.
>
> Kernel 2.6.26-2-xen-amd64, DRBD 8.3.5 compiled from Debian unstable package
> for 8.3.4
>
> For anyone interested, here is the stack trace.
> BR,
> Ivars
>
>
> Nov 16 03:00:29 ariel kernel: [31375.026193] BUG: unable to handle kernel
> NULL pointer dereference at 0000000000000016
> Nov 16 03:00:29 ariel kernel: [31375.026288] IP: [<ffffffffa02f9169>]
> :drbd:drbd_connector_callback+0x32/0x181
> Nov 16 03:00:29 ariel kernel: [31375.026359] PGD 164c4067 PUD 170d1067 PMD 0
> Nov 16 03:00:29 ariel kernel: [31375.026423] Oops: 0000 [1] SMP
> Nov 16 03:00:29 ariel kernel: [31375.026474] CPU 0
> Nov 16 03:00:29 ariel kernel: [31375.026512] Modules linked in: xt_physdev
> iptable_filter ip_tables x_tables sha1_generic dr
> bd cn iscsi_trgt crc32c libcrc32c ipv6 bridge xfs w83627ehf lm85 hwmon_vid
> netconsole configfs xenblktap netloop softdog ipm
> i_watchdog ipmi_msghandler loop psmouse serio_raw pcspkr i2c_i801 i2c_core
> button rng_core shpchp pci_hotplug intel_agp evde
> v ext3 jbd mbcache dm_mirror dm_log dm_snapshot dm_mod ide_cd_mod cdrom
> ide_disk ide_pci_generic ata_piix piix ide_core ata_
> generic libata scsi_mod dock skge ehci_hcd uhci_hcd thermal processor fan
> thermal_sys [last unloaded: scsi_wait_scan]
> Nov 16 03:00:29 ariel kernel: [31375.027370] Pid: 3165, comm: cqueue Not
> tainted 2.6.26-2-xen-amd64 #1
> Nov 16 03:00:29 ariel kernel: [31375.027405] RIP: e030:[<ffffffffa02f9169>]
> [<ffffffffa02f9169>] :drbd:drbd_connector_callb
> ack+0x32/0x181
> Nov 16 03:00:29 ariel kernel: [31375.027485] RSP: e02b:ffff8800104f3e50
> EFLAGS: 00010206
> Nov 16 03:00:29 ariel kernel: [31375.027519] RAX: 0000000000000000 RBX:
> ffff88001648c220 RCX: 0000000000000000
> Nov 16 03:00:29 ariel kernel: [31375.027555] RDX: 0000000000000000 RSI:
> 0000000000000000 RDI: ffff8800164c9c10
> Nov 16 03:00:29 ariel kernel: [31375.027597] RBP: ffff88001648c1d8 R08:
> ffff8800104f2000 R09: ffffffff80553e18
> Nov 16 03:00:29 ariel kernel: [31375.027633] R10: 0000000000000000 R11:
> 7fffffffffffffff R12: ffff8800164c9c10
> Nov 16 03:00:29 ariel kernel: [31375.027669] R13: ffffffffa02d30c3 R14:
> ffffffff8057d1c0 R15: 0000000000000000
> Nov 16 03:00:29 ariel kernel: [31375.027709] FS: 00007f9ee13c46e0(0000)
> GS:ffffffff8053a000(0000) knlGS:0000000000000000
> Nov 16 03:00:29 ariel kernel: [31375.027761] CS: e033 DS: 0000 ES: 0000
> Nov 16 03:00:29 ariel kernel: [31375.027793] DR0: 0000000000000000 DR1:
> 0000000000000000 DR2: 0000000000000000
> Nov 16 03:00:29 ariel kernel: [31375.027829] DR3: 0000000000000000 DR6:
> 00000000ffff0ff0 DR7: 0000000000000400
> Nov 16 03:00:29 ariel kernel: [31375.027866] Process cqueue (pid: 3165,
> threadinfo ffff8800104f2000, task ffff8800161e1440)
> Nov 16 03:00:29 ariel kernel: [31375.027918] Stack: 0000000000000000
> ffff88001648c220 ffff88001648c1d8 ffff88001648c1d0
> Nov 16 03:00:29 ariel kernel: [31375.028024] ffffffffa02d30c3
> ffffffff8057d1c0 0000000000000000 ffffffffa02d30d8
> Nov 16 03:00:29 ariel kernel: [31375.028120] 7fffffffffffffff
> ffff880016f76840 ffff88001648c1d0 ffffffff8023c34c
> Nov 16 03:00:29 ariel kernel: [31375.028185] Call Trace:
> Nov 16 03:00:29 ariel kernel: [31375.028250] [<ffffffffa02d30c3>] ?
> :cn:cn_queue_wrapper+0x0/0x33
> Nov 16 03:00:29 ariel kernel: [31375.028393] [<ffffffffa02d30d8>] ?
> :cn:cn_queue_wrapper+0x15/0x33
> Nov 16 03:00:29 ariel kernel: [31375.028439] [<ffffffff8023c34c>] ?
> run_workqueue+0xbe/0x189
> Nov 16 03:00:29 ariel kernel: [31375.028482] [<ffffffff8023cd35>] ?
> worker_thread+0xd5/0xe0
> Nov 16 03:00:29 ariel kernel: [31375.028522] [<ffffffff8023f6c1>] ?
> autoremove_wake_function+0x0/0x2e
> Nov 16 03:00:29 ariel kernel: [31375.028564] [<ffffffff8023cc60>] ?
> worker_thread+0x0/0xe0
> Nov 16 03:00:29 ariel kernel: [31375.028601] [<ffffffff8023f593>] ?
> kthread+0x47/0x74
> Nov 16 03:00:29 ariel kernel: [31375.028637] [<ffffffff802283a8>] ?
> schedule_tail+0x27/0x5c
> Nov 16 03:00:29 ariel kernel: [31375.028677] [<ffffffff8020be28>] ?
> child_rip+0xa/0x12
> Nov 16 03:00:29 ariel kernel: [31375.028722] [<ffffffff8023f54c>] ?
> kthread+0x0/0x74
> Nov 16 03:00:29 ariel kernel: [31375.028760] [<ffffffff8020be1e>] ?
> child_rip+0x0/0x12
> Nov 16 03:00:29 ariel kernel: [31375.028796]
> Nov 16 03:00:29 ariel kernel: [31375.028824]
> Nov 16 03:00:29 ariel kernel: [31375.028852] Code: 41 55 41 54 49 89 fc 55
> 53 48 83 ec 08 65 8b 04 25 24 00 00 00 83 3d a6 75 01 00 02 74 1e 89 c0 48
> c1 e0 07 48 ff 80 00 09 31 a0 <f6> 42 16 20 be 98 00 00 00 0f 84 20 01 00 00
> eb 1a 41 5b 5b 5d
> Nov 16 03:00:29 ariel kernel: [31375.029581] RIP [<ffffffffa02f9169>]
> :drbd:drbd_connector_callback+0x32/0x181
> Nov 16 03:00:29 ariel kernel: [31375.029657] RSP <ffff8800104f3e50>
> Nov 16 03:00:29 ariel kernel: [31375.029688] CR2: 0000000000000016
> Nov 16 03:00:29 ariel kernel: [31375.030762] ---[ end trace 296f6157c8798c56
> ]---
>
>
> Jean-Francois Chevrette wrote:
>>
>> It appears that there is currently a problem with the latest CentOS/Redhat
>> kernel. We have noticed the same problem when using LVM snapshots and a
>> backup technology called R1Soft CDP.
>>
>> Some related info:
>> http://bugs.centos.org/view.php?id=3869
>> forum.r1soft.com/showthread.php?t=1158
>>
>> No sign of a bug at bugzilla.redhat.com
>>
>> For now we have reverted to kernel-2.6.18-128.7.1 on which we did not have
>> any issues for the past 4 hours. Previously, a few seconds after starting a
>> 'drbdadm verify' the kernel panic would occur.
>>
>> DRBD devs might want to check it out.
>>
>> Regards,
>
> _______________________________________________
> drbd-user mailing list
> drbd-user@...
> http://lists.linbit.com/mailman/listinfo/drbd-user
>

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: Breaking DRBD pair?

2009-12-14T12:41:55Z

Hello again,

I see that if I provide a non-zero value for "wfc-timeout" in drbd.conf, the primary server will boot with all services running and the DRBD volume mounted. Perhaps not the best solution, but it gets the job done.

ltollefs wrote:

Hello,

I need to retire the DRBD secondary server, breaking the pairing, while not experiencing ANY down time on the primary. We do not anticipate replacing the secondary server.

AUTO: Halina Skora is out of the office. (returning 01/04/2010)

2009-12-14T03:00:55Z

I am out of the office until 01/04/2010.

If you need immediate assistance please contact Paul O'Brien
(pobrien@...) or Jerry Roybal (jroybal@...) otherwise
I will get back to you when I return.

Note: This is an automated response to your message "drbd-user Digest, Vol
65, Issue 27" sent on 12/14/2009 4:00:01 AM.

This is the only notification you will receive while this person is away.

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Breaking DRBD pair?

2009-12-13T09:02:23Z

Hello,

I have inherited a pair of production servers running Fedora Core 6 and DRBD 8.0.5. I have limited exposure to DRBD and HA so I want to tread carefully here.

I need to retire the DRBD secondary server, breaking the pairing, while not experiencing ANY down time on the primary. We do not anticipate replacing the secondary server.

My question is, what configs on the primary need to be altered so the the primary will run as any server would, but just happens to be running drbd. Can this be done?

Thanks!

Re: DRBD Ingres issue

2009-12-11T06:27:10Z

On Fri, Dec 11, 2009 at 10:23 PM, Mark Watts <m.watts@...> wrote:

>
>> >
>> It's easy if it's MySQL or PostgreSQL because you just move the
>> database location but it's not like that with Ingres, it's treated as
>> a third party software wherein the default install location is the
>> /opt directory. In my case instead of the default I installed Ingres
>> on the drbd partition (system files and database, it doesn't put files
>> on /etc).
>>
>> I'm using as guide the Ingres HA cluster documentation wherein my
>> shared drive is my drbd partition, the problem is I can't even get it
>> to work manually much more automatically.
>
> Well in theory, installing apps directly to the shared filesystem is
> better since you _know_ you have everything you need to start the app on
> the second node.
>
> I assume you can get it up and running on the Primary node?
>
> Mark.

It's running on the primary node, even after I unmounted it and
demoted it as secondary then revert it back as primary Ingres runs but
not on the secondary node.
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: DRBD Ingres issue

2009-12-11T06:23:58Z

> >
> It's easy if it's MySQL or PostgreSQL because you just move the
> database location but it's not like that with Ingres, it's treated as
> a third party software wherein the default install location is the
> /opt directory. In my case instead of the default I installed Ingres
> on the drbd partition (system files and database, it doesn't put files
> on /etc).
>
> I'm using as guide the Ingres HA cluster documentation wherein my
> shared drive is my drbd partition, the problem is I can't even get it
> to work manually much more automatically.

Well in theory, installing apps directly to the shared filesystem is
better since you _know_ you have everything you need to start the app on
the second node.

I assume you can get it up and running on the Primary node?

Mark.

--
Mark Watts BSc RHCE MBCS
Senior Systems Engineer, Managed Services Manpower
www.QinetiQ.com
QinetiQ - Delivering customer-focused solutions
GPG Key: http://www.linux-corner.info/mwatts.gpg

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

signature.asc (204 bytes) Download Attachment

Re: DRBD Ingres issue

2009-12-11T06:15:55Z

On Fri, Dec 11, 2009 at 6:00 PM, Mark Watts <m.watts@...> wrote:

> On Fri, 2009-12-11 at 13:25 +0800, jan gestre wrote:
>> Hi Guys,
>>
>> I'm trying to configure a HA cluster using DRBD and RHEL cluster
>> suite, I'm currently in the DRBD stage wherein I'm testing the manual
>> failover and I'm having problems/difficulties resolving this one.
>>
>> I've /dev/drbd0 on both node1 and node2 and it's mounted as primary on
>> node1, I was able to install Ingres 9.3 on node1 using /dev/drbd0 as
>> install location however when I try to mount it node2, I can't start
>> Ingres -->
>>
>> [ingres@node2 ~]$ ingstart
>>
>> Ingres/ingstart
>>
>> No Ingres servers have been configured to start up.
>>
>> I hope someone here can help me out.
>
>
> Are you running in Primary/Secondary mode or Primary/Primary?

node1 was running as Primary/Secondary before I demoted it as
Secondary/Secondary
>
> If the former (which is most common), you'd need to make sure you've
> flipped the mount point to node2 before you try anything.
>
I unmounted /dev/drbd0 on node1 after demoting it to secondary then
promoted node2 to Primary.

> Also, there's an assumption that anything not located on the shared
> volume (anything in /etc for example) is also installed on the second
> node.
>
>
> For example, when we do this with MySQL or PostgreSQL, we install the
> regular distro packages on both nodes, then manually move the database
> directory to the shared storage and symlink it on both nodes. The
> Secondary node will have a dangling symlink until the filesystem is
> mounted, but this is OK since heartbeat will manage mounting the
> filesystem and starting the database service in the correct order for
> us.
>

It's easy if it's MySQL or PostgreSQL because you just move the
database location but it's not like that with Ingres, it's treated as
a third party software wherein the default install location is the
/opt directory. In my case instead of the default I installed Ingres
on the drbd partition (system files and database, it doesn't put files
on /etc).

I'm using as guide the Ingres HA cluster documentation wherein my
shared drive is my drbd partition, the problem is I can't even get it
to work manually much more automatically.
>
> --
> Mark Watts BSc RHCE MBCS
> Senior Systems Engineer, Managed Services Manpower
> www.QinetiQ.com
> QinetiQ - Delivering customer-focused solutions
> GPG Key: http://www.linux-corner.info/mwatts.gpg
>
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

best way to achieve a load balancer with drbd + ocfs2 + openais

2009-12-11T05:55:41Z

Hello,

I had gone through the documentations. But I am not able to understand
the concept of dlm in case of the ocfs2 drbd.

My setup is a load balancer + fail over for OpenVZ vpss.

1. There are two physical servers
2. The drbd1 formatted with ocfs2 in two nodes and mounted at /vz
3. The two drbd devices are in dual primary mode.
4. One VPS will be running in each node and, if one fails the openais
+ pacemaker will fail over the VPS running there to another online
node.

1. The openais is communication through two interfaces. They are eth0 and bond0
2. But drbd is communicating only through bond0 which is the bonded
master of eth1 and eth2.

The openais is managing the two nodes and two VPSs ( one vps in one
node and another in another node ). The OCFS2 uses its own heartbeat.

3. drbd and ocfs2 are not added in crm.

Here I am confused and note sure, how to add the drbd + ocfs2 ( ocfs2
uses its own heartbeat ).

PROBLEM :

in case If there is any problem with the bonding then the drbd devices
will disconnect. But the openais will work through eth0 and since the
drbd + ocsf2 is not added in crm, crm will not fail over the VPS to
another node and the two VPSs will resume working as it is.

But after connecting the drbd again, then it will detect the split
brain and it will ask to discard data in any of the node. I want to
save the data, so need a solution to this problem.

NOTE : I am running centos 5, so there is no user space dlm available
as in suse.

--
Regards,
Unni
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-11T03:19:44Z

On Thu, 2009-12-10 at 16:58 +0100, Florian Haas wrote:

> Lars beat me to this but anyway...
>
> On 2009-12-10 13:23, Beck Ingmar wrote:
> >
> > Hi Florian,
> >
> > why improves bonding of the replication-connection not the performance.
>
> I did not say that. :) I said round-robin bonding _over 4 NICs_ would be
> detrimental to performance.
>
> > We currently have a direct connection for the 1 Gb DRBD replication and
> > I
> > wanted to activate an active-active bonding (with a further LAN-port). I
> > would expect nearly double of replication-performance.
>
> All of the following applies to DRBD Connected in protocol C:
>
> - With 1 GBit Ethernet link, you can do about 110 MB/s (provided your
> storage subsystem has that bandwidth);
>
> - With 2 GBit links bonded in round-robin mode, you top out at around
> 160-170 MB/s (even then you may have to tune kernel TCP parameters);
>
> - with 3 Gbit links, your throughput will drop down to around 110 MB/s;
>
> - with 4 links or more, you will get _significantly less_ throughput
> than a single unbonded link.
>
> Reason for all this: TCP reordering.
>
> > Or depends it on whether the connection between the nodes directly or
> > via 1-n switches.
>
> I am talking about back-to-back connections. Switches could make it worse.
>

Do Jumbo Frames make any difference here?

Mark.

--
Mark Watts BSc RHCE MBCS
Senior Systems Engineer, Managed Services Manpower
www.QinetiQ.com
QinetiQ - Delivering customer-focused solutions
GPG Key: http://www.linux-corner.info/mwatts.gpg

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

signature.asc (204 bytes) Download Attachment

Re: performance issues with DRBD :(

2009-12-11T02:34:44Z

Ok Florian,

I had misunderstood it.
So 2 NICs (as round-robin) are good ... but 3 or more NICs get no benefit or
degrade the performance, right?
And any statements relating to back-to-back connections.

Ok ... we use protocol C.

What kernel TCP parameters have to tune and what values should they get?

"Florian Haas" <florian.haas@...>
schrieb im Newsbeitrag news:4B211A89.5090500@......
> _______________________________________________
> drbd-user mailing list
> drbd-user@...
> http://lists.linbit.com/mailman/listinfo/drbd-user
>

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: DRBD Ingres issue

2009-12-11T02:00:43Z

On Fri, 2009-12-11 at 13:25 +0800, jan gestre wrote:

> Hi Guys,
>
> I'm trying to configure a HA cluster using DRBD and RHEL cluster
> suite, I'm currently in the DRBD stage wherein I'm testing the manual
> failover and I'm having problems/difficulties resolving this one.
>
> I've /dev/drbd0 on both node1 and node2 and it's mounted as primary on
> node1, I was able to install Ingres 9.3 on node1 using /dev/drbd0 as
> install location however when I try to mount it node2, I can't start
> Ingres -->
>
> [ingres@node2 ~]$ ingstart
>
> Ingres/ingstart
>
> No Ingres servers have been configured to start up.
>
> I hope someone here can help me out.

Are you running in Primary/Secondary mode or Primary/Primary?

If the former (which is most common), you'd need to make sure you've
flipped the mount point to node2 before you try anything.

Also, there's an assumption that anything not located on the shared
volume (anything in /etc for example) is also installed on the second
node.

For example, when we do this with MySQL or PostgreSQL, we install the
regular distro packages on both nodes, then manually move the database
directory to the shared storage and symlink it on both nodes. The
Secondary node will have a dangling symlink until the filesystem is
mounted, but this is OK since heartbeat will manage mounting the
filesystem and starting the database service in the correct order for
us.

Hope this makes sense?

Mark.

--
Mark Watts BSc RHCE MBCS
Senior Systems Engineer, Managed Services Manpower
www.QinetiQ.com
QinetiQ - Delivering customer-focused solutions
GPG Key: http://www.linux-corner.info/mwatts.gpg

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

signature.asc (204 bytes) Download Attachment

DRBD Ingres issue

2009-12-10T21:25:44Z

Hi Guys,

I'm trying to configure a HA cluster using DRBD and RHEL cluster
suite, I'm currently in the DRBD stage wherein I'm testing the manual
failover and I'm having problems/difficulties resolving this one.

I've /dev/drbd0 on both node1 and node2 and it's mounted as primary on
node1, I was able to install Ingres 9.3 on node1 using /dev/drbd0 as
install location however when I try to mount it node2, I can't start
Ingres -->

[ingres@node2 ~]$ ingstart

Ingres/ingstart

No Ingres servers have been configured to start up.

I hope someone here can help me out.

TIA
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

drbd over mdraid-10?

2009-12-10T18:02:50Z

Hi

What do you thing about building drbd over Linux soft raid10 partition?
Will it hurt too much on the performance side?

I have a 3ware controller and I'm really disappointed with it's random
I/O operations... I think it's the root of all my problems, and I'm
considering exporting all disks as JBOD's and trying with mdraid. I have
raid 10 on another machine, and performance simply rocks - vs the 3ware
(with cache enabled) offcourse.

So my question is - is it OK to use drbd over the soft raids?

--
| Jakov Sosic | ICQ: 28410271 | PGP: 0x965CAE2D |
=================================================================
| |

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-10T14:05:43Z

On Thu, 2009-12-10 at 19:11 +0100, Lars Ellenberg wrote:

> exactly same hardware?

yes

> exactly same kernel and stuff?

CentOS 5.4, I really look after details, and 'rpm -qa | sort' give me
exactly the same output on both nodes, config files are also same,
everything is OK.

I tried reversing roles, and when secondary becomes primary, then the IO
load reverses too.

I found on the internet that 3ware controllers have issues on Linux, and
I have 3ware controller... So maybe that is the part of problem...

> io scheduler? -> use deadline.

I'm using it... I know about kernel threads problem with CFQ.

> otherwise: we sell "DRBD Healthchecks" ;)

I'm just a contractor too, but I think that the owners will buy FC
storage and get on without cheap JBODs...

--
| Jakov Sosic | ICQ: 28410271 | PGP: 0x965CAE2D |
=================================================================
| |

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

drbd1 write speeds 30% of backing device (CentOS 5.4 drbd 8.3)

2009-12-10T10:27:22Z

Hi,

I’m attempting to build a redundant storage solution using drbd replication,

however I can't seem to get the performance I would have expected.

I have the two servers on an Infiniband link, which has tested at about 170MB/s

(Megabytes not Megabits) with netcat (CPU bound).

Both disks write at just under 100MB/s (tested via dd). Yet when attempting to

write to the drbd device on the primary I consistently get 30MB/s. I assume

the problem must be something I've configured incorrectly, but for the life of

me I can't figure out what it is.

My configuration can be found at: http://pastie.org/737624

I have tried Google, and read and played with the tuning recommendations

in the drbd manual, with no success. In an effort to demonstrate the

problem is not network speeds, or the secondary, I have brought down the

secondary, and still see the same performance:

[root@lager ~]# cat /proc/drbd

version: 8.3.2 (api:88/proto:86-90)

GIT-hash: dd7985327f146f33b86d4bff5ca8c94234ce840e build by mockbuild@..., 2009-08-29 14:07:55

1: cs:WFConnection ro:Primary/Unknown ds:UpToDate/DUnknown C r----

ns:1048576 nr:0 dw:0 dr:1048576 al:0 bm:64 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0

[root@lager ~]# dd if=/dev/zero of=/dev/drbd1 bs=512M count=2

2+0 records in

2+0 records out

1073741824 bytes (1.1 GB) copied, 32.6865 seconds, 32.8 MB/s

[root@lager ~]# drbdadm down r0

[root@lager ~]# dd if=/dev/zero of=/dev/sda2 bs=512M count=2

2+0 records in

2+0 records out

1073741824 bytes (1.1 GB) copied, 11.1106 seconds, 96.6 MB/s

If anyone has any suggestions as to how I can troubleshoot this further it

would be greatly appreciated.

Thanks!

Walton

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: performance issues with DRBD :(

2009-12-10T10:11:58Z

On Thu, Dec 10, 2009 at 05:00:05PM +0100, Jakov Sosic wrote:
> I've tried LACP too, and I have exact the same problems with drbd. When
> DRBD is disconnected, device gives me double to triple performance, also
> changing from protocol C to A or B gives no significant improvement.

Then maybe you are simply looking in the wrong direction.

exactly same hardware?
exactly same kernel and stuff?
io scheduler? -> use deadline.

otherwise: we sell "DRBD Healthchecks" ;)

--
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
__
please don't Cc me, but send to list -- I'm subscribed
_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Split Brain notification fails

2009-12-10T09:30:19Z

Hi all, this is my first post at DRBD list, my name is Andres and i'm noob at drbd.
I'm trying to install a HA Debian Lenny cluster with Pacemaker + DRBD.
Is almost done, so i am testing it. My problem is that when I forced a split brain, drbd didn't send a notification email as I configured at drbd.conf :
.....
    fence-peer "/usr/lib/drbd/crm-fence-peer.sh";
    split-brain "/usr/lib/drbd/notify-split-brain.sh andres@...";
}

If i try to run notify-split-brain.sh manually from command line I get the following error:
# sh /usr/lib/drbd/notify-split-brain.sh andres@...
      /usr/lib/drbd/notify-split-brain.sh: line 9: syntax error near unexpected token `>'
      /usr/lib/drbd/notify-split-brain.sh: line 9: `exec > >(2>&- ; logger -t "$PROG[$$]" -p local5.info) 2>&1'

I'm not a bash expert and line 9 is quite confusing for me.
Can someone help me at this point?
Thans in advance.
Andres.

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

Re: SLES 11 showstoppers in the drbd-8.3.6.tar.gz source distribution

2009-12-10T08:35:22Z

On Thu, 2009-12-10 at 08:18 -0800, Joe Bill wrote:

> Can also modify the following in 'drbd.spec.in', otherwise yast appears to look for packages by the name of chkconfig, instead of a single file...
>
> In 'drbd.spec.in', replace ...
>
> Requires(post): chkconfig
> Requires(preun): chkconfig
>
> with ...
>
> Requires(post): /sbin/chkconfig
> Requires(preun): /sbin/chkconfig
>
> and ...
>
> 'chkconfig --add drbd'
>
> with
>
> '/sbin/chkconfig --add drbd'
>

On Fedora/RHEL/CentOS etc there really is an RPM called "chkconfig", and
if you "Requires: <package>", dependency resolution will pull in the
correct package for you.

SuSE seem to call the package "aaa_base" which makes no sense at all to
me.

Does SuSE (or the others) pull in the correct package if you require
specific files instead of package names?

Mark.

--
Mark Watts BSc RHCE MBCS
Senior Systems Engineer, Managed Services Manpower
www.QinetiQ.com
QinetiQ - Delivering customer-focused solutions
GPG Key: http://www.linux-corner.info/mwatts.gpg

_______________________________________________
drbd-user mailing list
drbd-user@...
http://lists.linbit.com/mailman/listinfo/drbd-user

signature.asc (204 bytes) Download Attachment