[DRBD-user] DRBD locks up after PingAck did not arrive in time

[John Du]

Lars Ellenberg wrote:
> On Wed, Jan 14, 2009 at 11:50:52AM -0800, John Du wrote:
>   
>> We are running five pairs of DRBD servers.  Three pairs are on 32bit  
>> RHEL 4 and 2 pairs on 64bit RHEL5.  Saturday (January 3, 2008) night,  
>> there was some network problems and the two 64bit DRBD pairs locked up  
>> the primary severs and we had to power them off and on.  The 32 bit  
>> servers are all survived fine.  Is this a known problem in the 64bit  
>> versions of DRBD we are running?  Will upgrading to the latest version  
>> fix the problem?  Your help is greatly appreciated.
>>
>> The Linux kernel version is 2.6.18-8.1.15.el5 #1 SMP Thu Oct 4 04:06:39  
>> EDT 2007 x86_64 x86_64 x86_64 GNU/Linux and the DRBD version is 8.2.0-3.
>>     
>
> you are not serious?
> you don't use 8.2.0, do you?
>
> should that be a typo, and you actually use 8.3.0 ?
> if not, please upgrade.
>
>   

>> The logs from one of the locked up servers is:
>>     
>
>   
>> 22:47:57 drbd1: Handshake successful: Agreed network protocol version 87
>>     
>
> ah. no.
> you _do_ use 8.2.0. where did you get that from?
> why?
>
>   

I do not remember where I got it from but I am sure I did not make it up 
myself.  Why should it not be used? Too buggy?  I have never known 
that.  Thank you for the info.

> in that case:
> it is likely a race condition bug, and that it triggered with one of
> your clusters and not the other is unlikely to be architechture related.
>
> and yes, upgrading will probably fix it.
>
>
> also, please note,
> just in case you run into any such problems again,
>
>   
>> 22:48:34 ----------- [cut here ] ---------  
>>     
>
> why, oh why, did you not include this part?
> that is the most interessting one...
>
>   
>> [please bite here ] ---------
>>     
> That is all the log the machine had before it locked up.  So I have nothing more to give to you. Sorry.
>