hello,
we have been using squid-2.6.6 (with some security patches from higher
versions and the kevent patch from .11) for some time now but few weeks ago
we started getting:
2007/03/17 08:25:12| comm_select: kevent failure: (9) Bad file descriptor
2007/03/17 08:25:12| Select loop Error. Retry 1
this error is very infrequent and can appear randomly few times a day (1 to
maximum of 22 so far) or does not at all for a few days and is not closely
related to load or openfiles (meaning it never occured either at lowest or
highest load). since my knowledge of C is limited i am not going to make any
assumptions unless asked to do so.
we have been running with squid for more than 4 years now (switching
temporarily to oops for a while because of unefficient select() loop) and
our central proxy now handles around ~550req/s at peak hours (1800-2100)
with 80Mbit/s or more. squid is configured not to cache anything at all and
and cache_dir is set to null since saving bandwidth via caching is not the
issue here (was 2 years ago). we can't revert back to non-proxy as that
would require a reconfigure of several thousand clients (and we don't want
to because of fine acls squid provides). we are using ~200 class 2
delay_pools to limit clients at the application level. we tried switching
the load among several proxy servers but the error did not disappear so we
reverted back to central one backuped via carp.
squid -k reconfigure is run 2 times a day to adjust delay_pools bandwidth
levels.
system runs freebsd-6.2 (releng_6 at 15th of March 2007) with 1gb ram and
polling enabled to minimize the load (~20kpps) and since the introduction of
squid kevent polling it worked great.
as we failed to resolve this problem and can't possibly imagine how to debug
via gdb at this load we are kindly asking for your assistance.
regards,
s.
