Re: [squid-users] reading swap.state file

On 5/08/2013 10:45 p.m., Hussam Al-Tayeb wrote:
> On Monday 05 August 2013 10:12:41 Amos Jeffries wrote:
>> On 5/08/2013 1:11 a.m., Hussam Al-Tayeb wrote:
>>> how I can parse swap.state file for inconsistencies?
>>> i.e. files referenced in swap.state but not in disk cache.
>> swap.state is a journal of transactions. It includes references to
>> operations that occured on old deleted files as part of its normal
>> content. Squid handles any such files without the delete record
>> automatically you do not need to worry about it.
>>
>>> Or files on disk but not referenced in swap.state. it seems squid does not
>>> know how to shut down correctly if one of the users is viewing a youtube
>>> video.
>> What do you mean by that last one? If Squid is shutting down it waits
>> shutdown_timeout for clients to finish up (a long video would not do so)
>> then terminates all remaining client connections and stops. None of them
>> get written to the log, and on restart the corrupted files will be
>> overwritten.
>>
>> Amos
> That's not the case here. I have not set shutdown_timeout so it should be the
> default. now on kernel updates I need to restart the server.
>
> sometimes I get more files on disk that cache.log says (this very rarely
> happens).
>
> And sometimes cache.log claims that swap.state has more referenced files than
> actually on disk. This happens almost every single time I squid -k shutdown
> then restart my server. Access.log file says the last thing being downloaded is
> a youtube video.

My earlier description is what happens on shutdown. If you have not set
shutdown_timeout yourself then the default of 30 seconds will be when
the timeout stages happen that is all. The process *will* happen
regardless of when the timeout occurs. Before the timeout Squid is
simply waiting for existing clients to finish and clearing up as much
disk I/O as possible.
It is good to hear that the cache garbage is a rare event, that means
your Squid is getting enough cycles on shutdown to finish the disk I/O
erase events at least.

I expect that the swap.state claiming more files than exist is something
to do with your OS caching disk I/O.

Amos
Received on Mon Aug 05 2013 - 13:27:08 MDT