Re: problem with storeSwapOutFileOpened

"Javier Puche. CSIC RedIRIS" writes:

>Dear all,
>
> My squid cache (Squid 2.1PATCH2 with Henrik recomended patches running
>on Solaris 2.6, ultra 2 machine) is having trouble; median access times
>are getting to over 5 seconds (even for hits); for one thing it seems
>to be overloaded (2000 TCP per minute, 6000 ICP per minute), LRU last
>days was really poor (0.20 days for a 14GB cache) but today got to 0.00
>days, secondly I am alarmed by the following messages in cache.log:
>
> (2) No such file or directory
>1999/02/18 13:37:04| storeSwapOutFileOpened: Unable to open swapfile:
>/cache5�/00/00/00000038
>
> where /cache5p is not a directory, it should be (as defined in
>squid.conf) /cache5/..., this kind of messages are being repeated very
>very often. Does someone knows what might them be due to ??
>
> another bad symptom is that swap size reported by SNMP is cero while
>cachemgr.cgi reports correct 90% usage of disks.
>
> no messages at the console this time (as I used to get many when the
>UFS getting-full-too-early problem with Solaris)

I can't explain the "cache5p" other than it could be a memory
corruption bug.

If your cache is not able to keep up with the load, you might want
to set the 'max_open_disk_fds' option in squid.conf. This will
limit the number of disk files that get opened for writing. Setting
this limit will decrease your response times, but it will also
decrease your hit ratio.

Duane W.
Received on Thu Feb 18 1999 - 10:30:09 MST