[OpenAFS] OpenAFS 1.4.1 and Solaris 9 "hangs"

Jeff Blaine jblaine@mitre.org
Thu, 07 Sep 2006 15:44:50 -0400


Any ideas would be great.

We're experiencing odd "hangs" in vos commands.
-- Nothing logged to /usr/afs/logs/*
-- Nothing logged to syslog
-- Solaris 9 on all boxes in this testbed
-- OpenAFS 1.4.1
-- ping between all boxes is a constant 0ms
-- No router is involved

Running "apptrace vos examine u.jer", for example, worked fine
6 times in a row.  The 7th time resulted in a series of the
following blocks and took a full 1m45s to finish (successfully).
Note the "dangling" open paren after the call to
select_large_fdset.  This is where it hiccups for a second or
three before spitting out another big block of this.

...
vos      -> libc.so.1:gettimeofday(tp = 0xdab14, tzp = 0x0) = 0x0
vos      -> libc.so.1:gettimeofday(tp = 0xfae80, tzp = 0x0) = 0x0
vos      -> libc.so.1:gettimeofday(tp = 0xfae18, tzp = 0x0) = 0x0
vos      -> libc.so.1:malloc(size = 0x4) = 0xce390
vos      -> libc.so.1:.umul(0x2200, 0x10624dd3, 0xdba98)
vos      -> libc.so.1:.umul(0x2200, 0x10624dd3, 0xdba98)
vos      -> libc_psr.so.1:memset(s = 0xfaca0, c = 0x0, n = 0x10) = 0xfaca0
vos      -> libc_psr.so.1:memset(s = 0xdbb60, c = 0x0, n = 0x1c) = 0xdbb60
vos      -> libc_psr.so.1:memset(s = 0xfac00, c = 0x0, n = 0x18) = 0xfac00
vos      -> libsocket.so.1:sendmsg(s = 0x3, msg = 0xfac00, flags = 0x0) 
= 0x2c
vos      -> libc.so.1:time(tloc = 0x0) = 0x45006517
vos      -> libc.so.1:free(ptr = 0xce390)
vos      -> libc.so.1:gettimeofday(tp = 0xfae80, tzp = 0x0) = 0x0
vos      -> libc.so.1:time(tloc = 0x0) = 0x45006517
vos      -> libc_psr.so.1:memset(s = 0xfd0a0, c = 0x0, n = 0x34) = 0xfd0a0
vos      -> libc_psr.so.1:memset(s = 0xabb50, c = 0x0, n = 0x2000) = 0xabb50
vos      -> libc_psr.so.1:memset(s = 0xadb50, c = 0x0, n = 0x2000) = 0xadb50
vos      -> libc_psr.so.1:memset(s = 0xafb50, c = 0x0, n = 0x2000) = 0xafb50
vos      -> libc.so.1:gettimeofday(tp = 0xdab14, tzp = 0x0) = 0x0
vos      -> libc_psr.so.1:memset(s = 0xabb50, c = 0x0, n = 0x2000) = 0xabb50
vos      -> libc_psr.so.1:memset(s = 0xadb50, c = 0x0, n = 0x2000) = 0xadb50
vos      -> libc_psr.so.1:memset(s = 0xafb50, c = 0x0, n = 0x2000) = 0xafb50
vos      -> libc.so.1:select_large_fdset(