[ic] IC gone wild ?

Jon Jensen jon at endpoint.com
Sat Aug 22 21:39:17 UTC 2009


On Sat, 22 Aug 2009, DB wrote:

> I'm running 5.6.1 on a fairly beefy Centos 5.3 server. I have about 
> 650,000 items in a Mysql database. Things run very smoothly for the most 
> part.
>
> Yesterday I began getting complaints from employees and customers about 
> the site acting funny... they tell of seeing a blank "white screen" 
> appearing after attempting to search for an item, add an item to the 
> cart etc.
>
> Finally I noticed that my main IC (not catalog) error log was quite
> large. It was not getting rotated (my fault or course) and had grown to
> 131MB. In the log were thousands of lines like:
>
> - - - [22/August/2009:08:19:06 -0400] - - page server pid 25756 won't die!
> - - - [22/August/2009:08:19:08 -0400] - - Spawning 5 page servers to
> reach 5 StartServers
> - - - [22/August/2009:08:19:08 -0400] - - page server pid 25762 won't die!
> - - - [22/August/2009:08:19:08 -0400] - - page server pid 25761 won't die!
> - - - [22/August/2009:08:19:10 -0400] - - Spawning 5 page servers to
> reach 5 StartServers
> - - - [22/August/2009:08:19:10 -0400] - - page server pid 25768 won't die!
>
> Rotating the log and restarting IC seems to have cured things, but I'm
> wondering what caused this. Is it possible that the large size of the
> error log was to blame? Any thoughts are welcomed.

A 131 MB log file is probably totally unrelated. I've seen them grow to 
several GB without ill effect on Interchange. (Of course it's a good idea 
to rotate them for other reasons. :)

At first glance it sounds like an old PreFork race condition that can 
strike on really powerful hardware, but that was fixed in 2007 before 
Interchange 5.6.1 came out.

It could be as simple as some page having an infinite loop or taking too 
long and being killed by housekeeping, which would result in a "white 
page" to the user. That could result in the errors you saw in your log, of 
page servers that won't die and then being forcibly killed and new page 
servers being spawned to meet StartServers.

I would recommend you try to track down what people are searching for, 
check your MySQL logs to see how long the queries are taking to run, check 
your search results page to see if it's blowing up in memory using when 
processing the results, etc. It seems likely your problem will return 
unless you find the real cause. :(

Jon

-- 
Jon Jensen
End Point Corporation
http://www.endpoint.com/



More information about the interchange-users mailing list