Connect. Communicate. Collaborate. Securely.

Home » Kerio User Forums » Kerio Connect » Imbedded disk performance problem in Kerio! (Bug with status.fld unneeded writing in each touched by mail client folder.)
  •  
ros_tok

Messages: 10
Karma: 0
Send a private message to this user
Hello,

It looks that huge number of writes to disk with users messages are generated by "changing" status.fld files in corresponding mail folders, when user is accessing (or mail client check folder for modifications) any folder.

Can anyone help and explain what is the reason of such behaviour of kerio mailserver. Is there any possibility of disabling such action.

This behavious creates huge impact to performance of kerio mailserver. You have a lot of email clients scanning IMAP folders to check for modification. Each time folder is scanned, "status.fld" file in this folder changes modification file time. So system writes small data to disk. A lot of clients so create randomized trafic of small writes. In my case 50 clients generate a lot of such IOs. At 100 IOs/sec disk show 100% used time. So all is stuck into disk. No new real emails or anything normal. Only Idiotic writes created as result of folders scans. No real modifications in mail data base.

I found old topic where it was said to be corrected in version 6.3. ::))

Best Regards,
Igor Arsenin

[Updated on: Fri, 21 February 2014 06:05]

  •  
ros_tok

Messages: 10
Karma: 0
Send a private message to this user
Hello Kerio!

If anybody can comment this problem?

OK! I will try to describe situation in more details. Corresponds to real situation in our company.

1. We have Sales mail box, that contains a lot of letters in about 10000 IMAP folders. So many mail folders to sort customers.

2. 5 members of sales department work with this mail box. So at least 5 mail clients like MS Outlook and Live Mail periodically check all this folders for modifications.

3. During each folder check by mail client, Kerio Connect 8.2.2 rewrites file status.fld (WHY? Nothing changed in storage!). You can see it in File system directly and in Debug logs about Storage. This leads to few disk writes -- modification File System metadata and writing contents of status.fld itself. So lets say somethig like 5 small writes for each file.

4. So for one mailbox folders check by one client we have 10000*5 => 50000 small writes. That randomly placed on disk. So we have a lot of disk head movements.

5. According to statistic Our Mirrored disks can process about 100 such random writes pear second. Really good reasult. Each IO leads to disk head reposition.

6. So we have 50000 / 100 => 500 seconds for one client to scan folders. Or 100 * 5 clients => 5000 seconds for all our sales stuff. Check interval is shorter by default. So we have madly working Disks only for this mailbox and unhappy stuff who talking about very low mail performance.

7. We also have similar mailboxes and stuff of Technical Support!

8. Few personts of CPU usage by mail server. It stuck on DISK IO!

9. A few reads from disk needed for all of this above. Most data needed to check folder and see that nothing modified is in RAM (File Cache) already.

10. So I think without thouse PARASITIC WRITES of status.fld files. Scan of thouse BIG MailBoxes for each mail client will be few seconds on local network. And no needs for Grait Disk Arrays or SSD drives.

So I think that writes during Mail Folders check is BAD BUG in current version of Kerio connect. That leads to timeouts BUGs in mail processing -- sending, receiving. And to requests to significantly nore advanced DISK Subsystem than really needed.

Best Regards,
Igor Arsenin
  •  
Pavel Dobry (Kerio)

Messages: 5245
Karma: 251
Send a private message to this user
I would be helpful if you can provide more details, like what operating system you use and what filesystem.
  •  
Pavel Dobry (Kerio)

Messages: 5245
Karma: 251
Send a private message to this user
Thank you for reporting the issue. I can confirm it is happening with 8.2.x versions. We are working on a fix.
  •  
ros_tok

Messages: 10
Karma: 0
Send a private message to this user
Thanks! Smile

We use CentOS release 5.10 (Final), x86. Kerio Connect 8.2.2 No this problem were in 7.x versions, I think. In mount options set "noatime, nodiratime" to disable writing open file times. First impression was that this leads to parasitic writes. But real problem was different.

As you can see, no problems to reproduce this bug. You can see it simply in Debug storage log. Or start in mail client account synchronization and see that all status.fld files has new modification time.

We moved today evening mail store to SSD. So may be it will give to mail server more power against this bug.

Still it is Critical Performance BUG. That need to be corrected as fast as possible. Or you may lost part of company face.

Good Luck,
Igor Arsenin
Previous Topic: Public Contacts do not show up in CardDAV-Clients
Next Topic: Mac mail issues with Kerio
Goto Forum:
  


Disclaimer:
Kerio discussion forums are intended for open communication between forum members and may contain information and material posted by members which may be useful in learning about Kerio products. The discussion forums are not intended to provide technical support for any specific product. Any information implied or expressed in the discussion forums is that of the posting member. Kerio is in no way responsible for the information posted in the forums, or its accuracy. Kerio employees may participate in the discussions, but their postings do not represent an offical position of the company on any issues raised or discussed. Kerio reserves the right to monitor and maintain the forums to promote free and accurate exchange of information.

Current Time: Wed Aug 23 06:23:54 CEST 2017

Total time taken to generate the page: 0.00469 seconds
.:: Contact :: Home ::.
Powered by: FUDforum 3.0.4.