topic Re: perf issue with one of the nodes in StoreVirtual Storage

perf issue with one of the nodes

bublik — Tue, 13 Nov 2012 11:52:43 GMT

I had a performance issue where the entire cluster became degraded. Opened support ticket with HP but they failed to explain to me what is stored in store.info means. min_ticks=2 goes to very high number one node only. This is the exact node experience high IO when i logged on to CMC. Does anyone know how to explain this?

Re: perf issue with one of the nodes

Gediminas Vilutis — Tue, 13 Nov 2012 15:20:29 GMT

Can't tell much about store.info content. But in past I've experienced such behaviour when one or more disks on node started to fail. Ask support to check ADU report of degraded node - failing disk starts to log a lot of errors to this file.

Gediminas

Re: perf issue with one of the nodes

bublik — Tue, 13 Nov 2012 19:32:09 GMT

What is ADU report? is it only available to support via SSH?

Re: perf issue with one of the nodes

Gediminas Vilutis — Wed, 14 Nov 2012 07:51:28 GMT

It is a standard log file, you can take it from exported management group or node support bundle or simply download from node with CMC (under Diagnostics, Log files tab, hpadu.log and hpadu_xml.zip files).

Gediminas

Re: perf issue with one of the nodes

bublik — Wed, 14 Nov 2012 12:24:00 GMT

there is no ADU log

Also, do you know if it is a problem for a lefthand to have 2 clusters in one managaement group that have different network bonds setup? i am thinking it could be contributing to my issues lately. my first claster is 10G and second cluster is 1G type. 10G is using ALB and 1G using Link Aggregation Dynamic Mode (802.3ad). Some microsoft nodes talking to both clusters.

Re: perf issue with one of the nodes

David_Tocker — Thu, 22 Nov 2012 01:10:30 GMT

Nope network bonds can be different happily. LACP does tend to cause some funky issues with some switches however.