High server-load and very large temp-folder

[hs]

Since scalix 11 it seems to be necessary to use the newest hardware available !?
Every of our installation has an verry high load, somtimes 15 and more.
The processes are mostly
java
index.browse
in.imap41d
In meanwhile we did a lot to solve this, including reindexing - without success.
To make it a bit acceptable to the user we made script wich is renicing all the java and index.browse processes to 19! This is hard because they ar forking new childs permanently .
It seems to me that every incoming object is generating a new reindexing of the whole mailbox. Also the temp folder is a big problem: with more than 10000 files in it the system needs a while to accessing this folder.
A little bit helpful here is to clean all files older than - say 3 days.

But the problem is'nt solved.
So what i need to know: what are the processes wich are using this temp folder ?
Is there a way to put the indexer and, maybe some other processes, to another server to save the serverperformance for the ual-processes - because they are necessary to the users.

Strange is: this becomes more bad from 11.0.3 to 11.0.4

[hs]

Hi all,
is realy nobody here wich is familary with the internals ?
I could be interesting to know what realy happens when indexer is running ?
Is the clain above true ?
What is java doing the whole time ?
Can i switch off the indexer and the tomcat without having problems - only slow search ?
Is there realy no way to bring the cpu-hungry processes to another machine ?

[yuceli]

I didn't noticed/checked before reading this post. Yes, our temp folder also contains about 300.000 files (we have about 150 users). There are files dated from the first install of the system (March 2007), so it does not delete any files I guess. What are these files? How can we make them deleted after a period of time?

Yucel

[hs]

Hi Yucel,
Scalix recommends to delete the old files. There is no implied mechanism within Scalix today. You can safly use a cronjob wich runs i script like this:
#--------
find /var/opt/scalix/s1/s/temp/ -type f -ctime +3 |while read FN
do
rm -f $FN;
done
#--------
Anyway, this helps only a litle bit because the access to the folder is faster with less inodes - it's not a solution.

[hs]

This is strange:
on imap-process (from a notifylink-server) needs 100% cpu:

top - 22:07:51 up 19 days, 8:47, 2 users, load average: 1.10, 1.20, 1.09
Tasks: 265 total, 2 running, 263 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.2% us, 31.0% sy, 19.8% ni, 49.0% id, 0.0% wa, 0.0% hi, 0.0% si
Mem: 790776k total, 736084k used, 54692k free, 13832k buffers
Swap: 0k total, 0k used, 0k free, 114088k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
9047 60536 39 19 24380 5632 4060 R 100 0.7 21:48.30 in.imap41d
10739 root 17 0 3540 1068 760 R 2 0.1 0:00.33 top

Who've got an idea ?

[Shredder]

You can do a omshowu -U <userid> to see which user this imap process is associated with. Get the user to log out completely and see if it goes away. If not, you can kill -9 it or restart the imapd service with: omoff -d0 -w imap; omon imap

Shredder

[stio]

we had also a high server load.
After uninstalling SIS the load is constantly between 0.6 - 2.3 (average value 1.5).
Befor the uninstall of SIS we had a load between 3.8 - 15 (average value 6).