Re: Crash problems...

On Thu, 27 Jun 1996, Geosh Fathauer wrote:
>> 2) With squid 1beta 17, it will hang after a while. I have not
>> been able to track down the specific circumstances, but it will
>> stop serving requests after 24-48 hours.

Usually this seems to be around processing certian signals here...viz;

hangs after a kill -1 most times (reconfiguration seems to send the cache
into a loop often)

hangs after a kill -USR1 occasionally (we do these nightly, the higher
usage cache hangs during rotating logs every 5 days or so)

Check the logs and see if it hung while something like this was happening.
(eg: hung with logs half rotated)

Our smaller caches have been up well over a week running a local
version derived from 1.0beta17.

The version that has been most reliable is tagged here as "1.0beta17.UCS2"
(we redistribute versions which we find stable, with small log format
changes and any patches off the squid web site, to other local
universities).

Changes from beta17:

Patch "UCS": 2 fixes from http://www.nlanr.net/Squid/1.0beta17/ plus an
             addition to the httpd-format logs.

Patch "UCS1": patch "UCS" plus an OSF/1 specific bug fix (64-bit specific to
              be precise).

Patch "UCS2": patch "UCS1" plus changes from signal() to sigaction(); useful
              on several systems, important on OSF/1.

If you think this might be useful to you, we keep a copy on
ftp.uwa.edu.au:/pub/squid (there is a patch there from beta17 to
beta17.UCS2, as well as other assorted patches).

But I'm sure beta18 will fix most of this, I don't know what direction it
will take with sigaction tho. My "UCS2" changes the singal handlers in
the "squid" binary to use appropriate flags to sigaction, but I haven't
looked at and worked out which flags are appropriate to the handlers in
the other binaries (eg; ftpget still uses signal). This should only take
a minute or so if someone wants to do it, just that it isn't giving me any
problems...

David.
Received on Thu Jun 27 1996 - 19:32:41 MDT