High CPU load after upgrading to 9.2.1

[evydaemon]

I have a major problem with authpsa processes consuming a lot of cpu load.
Since the update to the new version it's extremely slow in the Plesk control panel and overall with web services...
Anyway, all services are up and running for now.

Update was finished without error messages.

I rebooted the server twice now:

#> top

Cpu(s): 2.3%us, 45.0%sy, 0.0%ni, 46.4%id, 6.2%wa, 0.0%hi, 0.1%si, 0.0%st
Mem: 2062284k total, 873684k used, 1188600k free, 129216k buffers
Swap: 3047416k total, 0k used, 3047416k free, 293052k cached

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3152 root 20 0 20384 1268 1072 D 11 0.1 2:48.81 authpsa
4023 root 20 0 20384 1264 1068 D 11 0.1 2:05.09 authpsa
4398 root 20 0 20384 1268 1068 R 11 0.1 1:49.24 authpsa
5098 root 20 0 20384 1264 1068 D 11 0.1 1:21.73 authpsa
6373 root 20 0 20384 1268 1068 D 11 0.1 0:29.31 authpsa
3740 root 20 0 20384 1268 1072 D 10 0.1 2:24.95 authpsa
4586 root 20 0 20384 1264 1068 D 10 0.1 1:38.78 authpsa
4731 root 20 0 20384 1264 1068 D 10 0.1 1:35.40 authpsa
4732 root 20 0 20384 1272 1068 D 10 0.1 1:33.98 authpsa
4988 root 20 0 20384 1268 1072 D 10 0.1 1:26.27 authpsa
3149 root 20 0 20384 1272 1068 D 10 0.1 2:49.35 authpsa
3689 root 20 0 20384 1272 1068 D 10 0.1 2:31.16 authpsa
4709 root 20 0 20384 1268 1068 D 10 0.1 1:36.51 authpsa
6736 root 20 0 20384 1268 1068 D 10 0.1 0:16.16 authpsa
3150 root 20 0 20384 1268 1068 R 10 0.1 2:47.39 authpsa
3151 root 20 0 20384 1264 1068 D 10 0.1 2:48.53 authpsa
3153 root 20 0 20384 1268 1072 D 10 0.1 2:49.44 authpsa
3154 root 20 0 20384 1272 1072 D 10 0.1 2:49.04 authpsa
3688 root 20 0 20384 1268 1068 D 10 0.1 2:30.08 authpsa
3890 root 20 0 20384 1264 1068 D 10 0.1 2:14.61 authpsa
4022 root 20 0 20384 1264 1068 D 10 0.1 2:06.32 authpsa
4024 root 20 0 20384 1268 1068 D 10 0.1 2:04.90 authpsa
5518 root 20 0 20384 1268 1068 D 10 0.1 1:00.52 authpsa
5633 root 20 0 20384 1268 1068 D 10 0.1 0:55.94 authpsa
5885 root 20 0 20384 1264 1068 D 10 0.1 0:45.64 authpsa
6540 root 20 0 20384 1264 1068 D 10 0.1 0:25.87 authpsa
3686 root 20 0 20384 1268 1068 D 9 0.1 2:30.61 authpsa
3714 root 20 0 20384 1268 1072 D 9 0.1 2:27.43 authpsa
4025 root 20 0 20384 1268 1068 D 9 0.1 2:05.25 authpsa
4453 root 20 0 20384 1272 1068 D 9 0.1 1:45.59 authpsa
4454 root 20 0 20384 1264 1068 D 9 0.1 1:45.20 authpsa
4455 root 20 0 20384 1268 1068 R 9 0.1 1:44.95 authpsa
4456 root 20 0 20384 1264 1068 D 9 0.1 1:44.32 authpsa

 

[hasher]

What OS? Try to switch off selinux/apparmor.

 

[manni]

Hi there,

same problem up here (see my thread)
Mail login is not working anymore on my machine.

Any suggestions?

 

[manni]

Debian 4.0

 

[Frank Hoffmann]

>> What OS? Try to switch off selinux/apparmor.

Same problem here ... server load is over 80 now, and update "hung" for 10 minutes, then continued.
The autoupdates tells me "selinux" is shut off.
But what is "apparmor." and how do I shut it down before an upgrade?

--->>>

07:39:33 up 1 day, 19:51, 2 users, load average: 83.46, 72.02, 49.31
USER TTY FROM LOGIN@ IDLE JCPU PCPU WHAT
root pts/0 c199077.adsl.han 06:54 40:06 19.38s 5.17s ./parallels_installer_v3.4.1_build090204.18_os_CentOS_5_i386
root pts/1 c199077.adsl.han 07:14 0.00s 0.01s 0.00s w

 

[evydaemon]

update:
Mail POP3 works if qmail service is stopped.
authpsa and relaylock processes don't appear if qmail is stopped via the panel.

 

[evydaemon]

fedora core 8 x64

fedora core 8 x64

 

[evydaemon]

Ok. Now there's an option under Home -> Settings -> Server-wide mail preferences called:

#Relaying:
authorization is required: POP3 lock time __Min.

if this is unchecked everything works fine. (of course SMTP authentication has to be activated to get mail from the server (same login as pop))
if checked and SMTP Service is running everything stops working after a while (relaylock processes flooding)

.. i tried to locate apparmor but it seems it's not installed. i looked in /etc/init.d and searched it with #>locate

 

[Frank Hoffmann]

u p d a t e :

upgrade finnshed finally -- server load went down to 1 or 2 AFTER the upgrade had finnished. Ten minutes after that upgrade I rebooted the server .. and boing! Now, after server ccomes back up the server load is at 70, then 80, now 103 and growing ... waiting for the crash. I see nothing but "relaylock" and "authpsa" prcesses.

This is a CentOS 5.3 server -- and it was a fresh Plesk 9.0.1 install that was now upgraded to 9.2.1.

I cna only warn anybody to upgrade! .... Anyone has some input here?

 

[manni]

hello
´
i found a good solution: change your mta with autoinstaller to postfix and it will working better then.
you may try a change to qmail back then, and this will work too I think.

plesk 9.2.1 is cool; no bugs until now, only 10 hours of getting crazy with the updates

manuel

 

[Benjamin Tauler Coma-Cros]

Wonder why did they call it 9.2.1

May be 9.2.1 looks stable? O_O

Guys, allways try Parallels Plesk updates on non production servers. Don't trust Parallels Plesk, they do not beta test anything. Just wait for them to correct the errors.

This was suposed to be the 9.1 version, but they named it 9.2.1 , may be because people was scared of using 9.0.1, to make more people to "test" its product... they named it that way, wich could look more stable.

The thing is that, untill they do not make like 4 or 5 revisions (no matter what name they have), the product is NOT STABLE. So keep that in mind.

Hope you guys can solve that problems you have right now.

Paralllels, you have a great product, but , GOD!!!, TEST YOUR PRODUCTS before publishing them.

 

[hasher]

Guys, DO NOT UPGRADE via apt-get/dpkg/yum/rpm/etc but via plesk autoinstaller! When we upgrade Plesk via package manager, it doesn't upgrade some packages, like a psa-mail-qc|pc-driver.

 

[Alexander Stamatis]

Upgraded via yum.

Everything worked perfectly ! Plesk upgrade took 3 mins without hangs etc.

------------

Yeap plesk needs time to mature but with 9.2.1 we had nothing to loose compared to 9.0.1

 

[evydaemon]

done it with /usr/local/psa/admin/sbin/autoinstaller -> selected postfix
everything works again.

after that and a reboot i ran autoinstaller again and sucessfully uninstalled qmail.

 

[Antonio Sanchez]

I have upgraded from 9.2.1 to 9.2.1.... (not an error) this morning using autoinstaller and the system boots ok.

Aprox. 30 minutes later I have experienced Heavy Load.

I had around 400 instances of 10-apam-XXXX/executable running. After a killall I got 1200 instances of relaylock.

When the server reaches 800,00 load average at one minute I performed a reboot.

After reboot I disabled grey listing spam features and the system comes to 4 to 5 load average over an hour. This is too high.

I disabled the pop3 session timeout before SMTP and the system comes down to 0.3 load average that is normal values.

Greeted for the solution.

PLEASE PARALLELS TEAM... Take a look to the Forum Usually....

 

[ZopfWare]

Only fix I've found is to roll back via rpm -Uhv with --force option to the 9.0.1 qmail files (both common driver and the qc)

Seems to be working as flawed as that version was, however no runaway authpsa and relyloick problems.

 

[mparadis]

It's getting worse all the time with parallels lately. Anyone have any sense of how quickly a fix is going to be made available? People can't even get at their email.

 

[mparadis]

Again, censoring posts.

 

[mparadis]

Under relay options, POP3 auth turned off keeps the load way down but what problems might this bring on?

 

[ZopfWare]

Has anyone tried my suggestion of rolling back to the 9.0.1 qmail rpms?? It worked for me and I'm no longer experiencing high load and I HAVEN'T had to disable any of the security features for my email.