You are viewing a plain text version of this content. The canonical link for it is here.
Posted to asp@perl.apache.org by "Mark T. Dame" <md...@mfm.com> on 2002/09/17 21:42:17 UTC
Problems with the State Manager
Hello all.
We have a strange problem with one of our web sites that uses
Apache::ASP. Every week or so it starts slowing down to a crawl. If we
clear out the StateDir, it returns to normal. It almost seems like
Apache::ASP isn't cleaning up it's state files under some circumstance
(although what those are is still a mystery).
We have also noticed errors like this:
[Mon Sep 16 21:06:48 2002] [error] Can't use string
("ile','refresh_timeout' =>
'7200'") as a HASH ref while "strict refs" in use at
/usr/lib/perl5/site_perl/5.
6.0/Apache/ASP/StateManager.pm line 232.
We are currently using Apache::ASP version 2.39. We have StateDB set to
DB_File. (Apache 1.3.26, mod_perl 1.27, perl 5.6.0)
Any ideas?
-m
--
## Mark T. Dame, Vice President, Internet Operations
## MFM Communication Software: http://www.mfm.com/
## E-mail: mailto:mdame@mfm.com WWW: http://www.mfm.com/~mdame/
"Yesterday I heard a tree muttering how it wished the HP4 would go
*up* [in price]."
-- Daniel Mocsny
---------------------------------------------------------------------
To unsubscribe, e-mail: asp-unsubscribe@perl.apache.org
For additional commands, e-mail: asp-help@perl.apache.org
Re: Problems with the State Manager
Posted by Josh Chamas <jo...@chamas.com>.
Mark T. Dame wrote:
>
>
> Yeah. I forgot to mention that part. I have all of my web directories
> (including the state directories) on NFS mounts since any given request
> can be handled by one of a number of physical servers.
>
Keep SDBM_File then, as its the most NFS corruption tolerant.
You need to make sure that your NFS clients & servers are supporting
flock() correctly over the network, or it really not safe to run the
StateDir in this way, though it mostly works as you have found.
Also, make sure your NFS clients never cache or buffer the data,
then you will also have data sync issues.
To make sure your NFS are locking correctly, try this test,
common StateDir over NFS between 2 servers. Have an ASP
script write something like:
<%= sprintf("%06d", $Application->{Count}++ )%>
Then hit the same script on 2 different servers with ab
for something like 1000 requests. The final result should
be something near 2001 when you view it in a browser, if
its not, you do not have flock() locking supported
across the network. I believe old NFS v2 did not do this
well, but that NFS v3 might, but have no real world experience
to back this up, just 2nd hand info.
If you want to continue with a StateDir approach, & NFS
does not get locking to work, then consider a samba mount.
Samba does support flock() locking on access to a network share.
If you have a NetApp, this might be a CIFS share instead.
If you want to abandon StateDir entirely, it is trivial to
wire up a custom $Session from an Apache::Session in Script_OnStart
for your application. What you gain in RDBMS backed sessions,
but you lose the session management & expiration.
Regards,
Josh
________________________________________________________________
Josh Chamas, Founder phone:925-552-0128
Chamas Enterprises Inc. http://www.chamas.com
NodeWorks Link Checking http://www.nodeworks.com
---------------------------------------------------------------------
To unsubscribe, e-mail: asp-unsubscribe@perl.apache.org
For additional commands, e-mail: asp-help@perl.apache.org
Re: Problems with the State Manager
Posted by "Mark T. Dame" <md...@mfm.com>.
Josh Chamas wrote:
>
> Mark T. Dame wrote:
> > Hello all.
> >
> > We have a strange problem with one of our web sites that uses
> > Apache::ASP. Every week or so it starts slowing down to a crawl. If we
> > clear out the StateDir, it returns to normal. It almost seems like
> > Apache::ASP isn't cleaning up it's state files under some circumstance
> > (although what those are is still a mystery).
> >
> > We have also noticed errors like this:
> >
> > [Mon Sep 16 21:06:48 2002] [error] Can't use string
> > ("ile','refresh_timeout' =>
> > '7200'") as a HASH ref while "strict refs" in use at
> > /usr/lib/perl5/site_perl/5.
> > 6.0/Apache/ASP/StateManager.pm line 232.
> >
> > We are currently using Apache::ASP version 2.39. We have StateDB set to
> > DB_File. (Apache 1.3.26, mod_perl 1.27, perl 5.6.0)
> >
>
> Looks like some kind of data corruption for the internal
> state database that would prevent Apache::ASP from cleaning
> up the state files correctly. Do you have the StateDir shared between
> hosts on some network file system?
Yeah. I forgot to mention that part. I have all of my web directories
(including the state directories) on NFS mounts since any given request
can be handled by one of a number of physical servers.
> This could cause corruption.
So SDBM_File doesn't handle NFS well?
> Failing that, you may have uncovered a scalability issue
> with SDBM_File which is used for the internal state dbm always
> for speed.
>
> If you want to change the type of dbm used for the internal
> database, you can do this:
>
> PerlModule Apache::ASP
> <Perl>
> $Apache::ASP::DefaultStateDB = 'DB_File';
> </Perl>
>
> StateDB only works on $Application & $Session, but not
> the internal database, so I believe the above will work.
I'll give this a shot and see what happens.
> I could also add some code that could work around the
> data corruption issue you are facing, and we can try
> that if you like.
We'll try the different internal state DB first. I'll let you know what
happens.
-m
--
## Mark T. Dame, Vice President, Internet Operations
## MFM Communication Software: http://www.mfm.com/
## E-mail: mailto:mdame@mfm.com WWW: http://www.mfm.com/~mdame/
"Where a calculator on the ENIAC is equipped with 18,000 vacuum
tubes and weighs 30 tons, computers in the future may have only
1,000 vacuum tubes and perhaps weigh 1 1/2 tons."
-- Popular Mechanics, March 1949
---------------------------------------------------------------------
To unsubscribe, e-mail: asp-unsubscribe@perl.apache.org
For additional commands, e-mail: asp-help@perl.apache.org
Re: Problems with the State Manager
Posted by Josh Chamas <jo...@chamas.com>.
Mark T. Dame wrote:
> Hello all.
>
> We have a strange problem with one of our web sites that uses
> Apache::ASP. Every week or so it starts slowing down to a crawl. If we
> clear out the StateDir, it returns to normal. It almost seems like
> Apache::ASP isn't cleaning up it's state files under some circumstance
> (although what those are is still a mystery).
>
> We have also noticed errors like this:
>
> [Mon Sep 16 21:06:48 2002] [error] Can't use string
> ("ile','refresh_timeout' =>
> '7200'") as a HASH ref while "strict refs" in use at
> /usr/lib/perl5/site_perl/5.
> 6.0/Apache/ASP/StateManager.pm line 232.
>
> We are currently using Apache::ASP version 2.39. We have StateDB set to
> DB_File. (Apache 1.3.26, mod_perl 1.27, perl 5.6.0)
>
Looks like some kind of data corruption for the internal
state database that would prevent Apache::ASP from cleaning
up the state files correctly. Do you have the StateDir shared between
hosts on some network file system? This could cause corruption.
Failing that, you may have uncovered a scalability issue
with SDBM_File which is used for the internal state dbm always
for speed.
If you want to change the type of dbm used for the internal
database, you can do this:
PerlModule Apache::ASP
<Perl>
$Apache::ASP::DefaultStateDB = 'DB_File';
</Perl>
StateDB only works on $Application & $Session, but not
the internal database, so I believe the above will work.
I could also add some code that could work around the
data corruption issue you are facing, and we can try
that if you like.
Regards,
Josh
________________________________________________________________
Josh Chamas, Founder phone:925-552-0128
Chamas Enterprises Inc. http://www.chamas.com
NodeWorks Link Checking http://www.nodeworks.com
---------------------------------------------------------------------
To unsubscribe, e-mail: asp-unsubscribe@perl.apache.org
For additional commands, e-mail: asp-help@perl.apache.org