[OpenAFS] repeated message: Delete longest inactive host

[Michal Svamberg]

Hello,

AFS Fileserver (OpenAFS 1.4.1 built  2006-05-05) was coming to meltdown slowly.
After debuging (kill -TSTP) to the fileserver on level 1,  it writes a
log message:

---cut---
Fri Nov  3 08:26:50 2006 [16] GSS: First looking for timed out call
backs via CleanupCallBacks
Fri Nov  3 08:26:50 2006 [16] GSS: Try harder for longest inactive host cnt= 1
Fri Nov  3 08:26:50 2006 [16] GSS: Try harder for longest inactive host cnt= 2
Fri Nov  3 08:26:50 2006 [16] GSS: Delete longest inactive host 147.228.53.104
... AND REPEATING THE SAME LINES ...
---cut---

During twenty seconds the fileserver produces 50MB of log (the same
lines above).
I try making a dump, but the fileserver make only empty files,and it
writes to FileLog:

---cut---
Fri Nov  3 08:32:16 2006 Created client dump
/etc/openafs/server-local/client.dump
Fri Nov  3 08:32:16 2006 Vice was last started at Fri Oct 20 08:32:54 2006

Fri Nov  3 08:32:16 2006 Large vnode cache, 600 entries, 20301 allocs,
124772344 gets (3140082 reads), 2808362 writes
Fri Nov  3 08:32:16 2006 Small vnode cache,600 entries, 304607 allocs,
82648006 gets (17546492 reads), 2655653 writes
Fri Nov  3 08:32:16 2006 Volume header cache, 600 entries, 125656518
gets, 584083 replacements
Fri Nov  3 08:32:16 2006 Partition /vicepa: 303787844 available 1K
blocks (minfree=0), Fri Nov  3 08:32:16 2006 239651500 free blocks
Fri Nov  3 08:32:16 2006 Partition /vicepb: 292960332 available 1K
blocks (minfree=0), Fri Nov  3 08:32:16 2006 257618224 free blocks
Fri Nov  3 08:32:16 2006 Partition /vicepc: 292960332 available 1K
blocks (minfree=0), Fri Nov  3 08:32:16 2006 255503768 free blocks
Fri Nov  3 08:32:16 2006 With 120 directory buffers; 10025532 reads
resulted in 212317 read I/Os
Fri Nov  3 08:32:16 2006 Total Client entries = 462, blocks = 265;
Host entries = 150, blocks = 1
Fri Nov  3 08:32:16 2006 There are 462 connections, process size 135544
Fri Nov  3 08:32:16 2006 There are 150 workstations, 20 are active
(req in < 15 mins), 1 marked "down"
Fri Nov  3 08:32:16 2006 Shutting down file server at Fri Nov  3 08:32:16 2006
Fri Nov  3 08:32:16 2006 Vice was last started at Fri Oct 20 08:32:54 2006

Fri Nov  3 08:32:16 2006 Large vnode cache, 600 entries, 20301 allocs,
124772371 gets (3140090 reads), 2808362 writes
Fri Nov  3 08:32:16 2006 Small vnode cache,600 entries, 304607 allocs,
82648008 gets (17546493 reads), 2655653 writes
Fri Nov  3 08:32:16 2006 Volume header cache, 600 entries, 125656545
gets, 584083 replacements
Fri Nov  3 08:32:16 2006 Partition /vicepa: 303787844 available 1K
blocks (minfree=0), Fri Nov  3 08:32:16 2006 239651500 free blocks
Fri Nov  3 08:32:16 2006 Partition /vicepb: 292960332 available 1K
blocks (minfree=0), Fri Nov  3 08:32:16 2006 257618224 free blocks
Fri Nov  3 08:32:16 2006 Partition /vicepc: 292960332 available 1K
blocks (minfree=0), Fri Nov  3 08:32:16 2006 255503768 free blocks
Fri Nov  3 08:32:16 2006 With 120 directory buffers; 10025532 reads
resulted in 212317 read I/Os
Fri Nov  3 08:32:16 2006 Total Client entries = 463, blocks = 265;
Host entries = 150, blocks = 1
Fri Nov  3 08:32:16 2006 There are 463 connections, process size 135544
Fri Nov  3 08:32:16 2006 There are 150 workstations, 20 are active
(req in < 15 mins), 1 marked "down"
Fri Nov  3 08:32:16 2006 VShutdown:  shutting down on-line volumes...
---cut---

These lines are writen to FileLog after  the Fileserver shutdown (in
lines in log show the same time as in shutdown).

Thanks for any ideas,
Michal Svamberg