[Ocfs2-users] OCFS2 tuning, fragmentation and localalloc option. Cluster hanging during mix read+write workloads

[Goldwyn Rodrigues]

Hi Gavin,

On 08/06/2013 04:51 PM, Gavin Jones wrote:
> Hello Goldwyn,
>
> Thanks for taking a look at this.  So, then, it does seem to be DLM
> related.  We were running fine for a few weeks and then it came up
> again this morning and has been going on throughout the day.
>
> Regarding the DLM debugging, I allowed debugging for DLM_GLUE,
> DLM_THREAD, DLM_MASTER and DLM_RECOVERY.  However, I don't see any DLM
> logging output in dmesg or syslog --is there perhaps another way to
> get at the actual DLM log?  I've searched around a bit but didn't find
> anything that made it clear.

Unfortunately CONFIG_OCFS2_DEBUG_MASKLOG is not enabled for opensuse 
kernels but for SLES kernels only. Sorry about that :(

However, you can recompile the kernel with this enabled in the config file.

>
> As for OCFS2 and iSCSI communications, they use the same physical
> network interface but different VLANs on that interface.  The
> "connectionX:0" errors, then, seem to indicate an issue with the ISCSI
> connection.  The system logs and monitoring software don't show any
> warnings or errors about the interface going down, so the only thing I
> can think of is the connection load balancing on the SAN, though
> that's merely a hunch.  Maybe I should mail the list and see if anyone
> has a similar setup.

You will not have anything in the logs if the network issues are 
intermittent. Perhaps a simple ping when the issue is occurring is the 
best tool.

My doubts on network issues keep getting stronger by the information you 
have given me so far. Also, as Sunil mentioned, you have a problem if 
the storage does not respond anyways.

>
> If you could please point me in the right direction to make use of the
> DLM debugging via debugs.ocfs2, I would appreciate it.
>

<snipped>

-- 
Goldwyn