Re: Problems witn memory in Squid 1.1.20 & FreeBSD

Andre Nuno Soares wrote:

> I have run into a problem with Squid 1.1.20 running in FreeBSD 2.2.5,
> and also in 2.2.2:
>
> When squid's process size gets to more or less 30M, the thing dies
> with one of 3 messages:
>
> FATAL: xcalloc: Unable to allocate 4096 blocks of 1 bytes!
> FATAL: xmalloc: Unable to allocate 4536 bytes!
>
> This happens on machines with 128 and 192 M of RAM, respectively, and
> when top and vmstat both indicate i still have memory (30M used for
> Cache...)
> I read a very small and vage post on dejanews about using GNU libmalloc
> when this happens, so could this be caused by the xcalloc and xmalloc?
> The first error messages is very odd, and looks to me like a bug...
>
> I apreciate all the help anyone can give me, since this is a very
> urgent problem (I only detect this when the servers went into
> production :((( )
>

Hi, I got the same error message before ( FreeBSD 2.2.5, squid 1.1.20,
128 MB RAM, cache_mem 45, Pentium II 266 MHz). according to mail
archive and the FAQ, I just do the two steps bellow to get rid
off the error:
1. edit /etc/login.conf, increase the daemon datasize
   (the default is 32MB)
2. recompile the kernel, with options:
   options "MAXDSIZ=(128*1024*1024)"
   options "DFLDSIZ=(128*1024*1024)"
   maxusers 40 ( I do this since the kernel always produces error
  "increase maxusers". fyi, in the peak hours we get about
  10 conn/second).

the error message dissapear but squid is eating all the availaible
memory which is the reason that I cron the daemon to shutdown
every 2 days to release the memory.

engel.
Received on Wed Jan 28 1998 - 08:22:19 MST