From: Steve D. <st...@us...> - 2005-11-11 23:49:28
|
"Thomas Guyot-Sionnest" <Th...@za...> wrote on 11/11/2005 04:55:48 PM: > Hi, Steve > > I found a fix for the CCM plugin, it's in CVS somewhere near v1.99.1 (CVS > date: "2005-02-02 19:38:37 +0000"). I made a patch out of it and applied it > to a fresh checkout of the STABLE_1_2 branch. It patched and compiled > cleanly, and fixes the quorum problem with ccm. > > Now, as soon as the STONITH occur, CCM gets the quorum (Verified with > evmsccm -v). This doesn't change much on evms's side.. > > It however breaks the failover. It also partially breaks the workaround with > ccm (starting ccm a few seconds after heartbeat; see > http://lists.community.tummy.com/pipermail/linux-ha-dev/2004-December/009685 > .html). When I use this workaround, I get a stale "evms -m > READWRITE_CRITICAL -b" process that blocks other evmd_failover scripts. If I > kill it, it works. > > That's not the case with the normal failover, which doesn't work at all. > > Do you have any idea of what's going wrong? Hello, Thomas. Are you sure the "evms -m READWRITE_CRITICAL -b" process is stale? It may take a few moments to complete. If the process is not completing, then please do this for me: 1. Edit /etc/evms.conf. In the "engine" section, set the "debug_level" to "debug". 2. Rerun the test. 3. When the evms process is stuck, get a copy of /var/log/evms-engine.log. gzip it and email it to me. Hopefully it will have some clues as to why it is not finishing. Steve D. |