Re: problem with squid (fwd)

Markus,

After I did kill -HUP squidid, it is wring fine until now ( about 12
hours). I still keep looking at it. Hope no problem occured again.

From start, I have not change any configuration. But our all the squid
software and caches on the disk shared by other machine. I had stopped
the squid run in original machine and started it on another machine. When
the original machine running, I had restart squid on original one again.
( of course, I stop the other machine first ). Than the problem are
occurring all the time through yesterday. I don't know about today.

I have to say it is a very painful to remove all the cache out and
restart again. Do I have any choice alse?

Best regards

Julie

On Wed, 29 Jul 1998, Markus Stumpf wrote:

> On Wed, Jul 29, 1998 at 04:38:17PM +1100, Julie Xu wrote:
> > Michael and experts,
> >
> > Bad new again, from cache.log I start to get message as below:
> >
> > 98/07/29 16:22:06| WARNING: Repeated failures to free up disk space!
> > 98/07/29 16:22:06| storeGetSwapSpace: Disk usage is over high water mark
> > 98/07/29 16:22:06| --> store_swap_high = 72181 KB
> > 98/07/29 16:22:06| --> store_swap_size = 268241 KB
> > 98/07/29 16:22:06| --> asking for 0 KB
> >
> > this message come out every minutes.
> >
> > now the cache is not full
> > sp3#df /cache
> > Filesystem 512-blocks Free %Used Iused %Iused Mounted on
> > sp1:/cache 1458176 727472 51% - - /cache
> >
> >
> > What is problem again???
>
> Ok, from the beginning:
>
> squid-1.1.x has a kind of "database" where it stores all the information
> about it's objects. This database is kept for one in memory and for
> two on disk. The latter is the file "log". When u send a kill -USR1 to
> squid, it dumps it's memory database.
> While squid is running it appends new fetched objects simply to the end
> of "log".
> If squid dies without a chance to write the clean "log" file it reads
> the "unclean" one and verifies it against the content of the cache
> directory(s). This is "slow startup".
> If it has a clean log file at startup (in this case it has a file
> "log-last-clean") it simply reads the file and no checks against
> the cache directories are done.
> In both cases it maintains a file "low.new" during startup and when
> it is finished it removes the "log" and renames "log.new" to log.
>
> The worst case that can happen to squid-1.1.x is that it loses the
> "log" file and has no chance to dump a clean new one.
> Then it doesn;t have any information on the content of the cache
> directories and all the objects are still there but squid is "blind"
> for them. This behaviour is fixed in 1.2.x
> In this situation you are probably running into a situation where
> your disk is physically full, but squid doesn't "know" about it.
> The clean way is to remove all the cache files/directories and
> "start over".
>
> The situation you have is very strange.
> Did you change something in the squid.conf? Maybe change the user
> squid is running under? In this case squid might not be able to
> unlink already existing (and outdated) files in the cache.
> >From the messages you receive I assume the latter in conjunction
> with a modification of the squid.conf file giving it less disk space
> to use.
> Squid does not simply "fill the disk until it's full" but you have
> to configure the amount of disk space squid may use.
>
> Hope that helps
>
> \Maex
>
Received on Wed Jul 29 1998 - 16:12:54 MDT