[OpenAFS] series of 1.4.2 fileserver crashes in rxi_FreeDataBufsTSFPQ (rx_packet.c)
Tue, 05 Dec 2006 09:51:25 +0100
We're having a series of fileserver crashes in rxi_FreeDataBufsTSFPQ,
about one a day, on a particular server running 1.4.2.
I looked at 6 dumps, the traceback is not always the same, but so far
the common clue is that call->conn->peer->host points to old afs 3.4
Never noticed this on 1.4.1, so a recent change might be responsible.
There have been a few deltas in rx, one dealing with congestion?!
Obviously I'll dig further into this, but as this is getting really
ugly now any quick idea is welcome (hence double-post, sorry!).
Downgrading to 1.4.1 has other ill-effects and is therefore only last
resort should identifying the volume(s) involved and moving them to an
old server fail.
European Laboratory for Particle Physics(CERN) - Geneva, Switzerland
Phone: +41 22 767 8985 Fax: +41 22 767 7155