Home » Kerio User Forums » Kerio Connect » Is Kerio capable to run over 350 users? (Having more issues ever since we went beyond 300 users)
  •  
Jorge

Messages: 20
Karma: 0
Send a private message to this user
Just want to know if anyone out there has experience running kerio connect for more than 300 accounts, it seems to me that ever since we went over 300 accounts the services is really bad really really slow... outlook (with outlook connector) remains unresponsive for a long time and more users complaint about the same issue.
We are running:

Windows Server 2012
16 cpus
32gb of ram
2.5tb of storage
on a VMWARE stack.

Kerio version 9.2.5 patch 3

I am just thinking maybe its time to move out of kerio?
Or what its wrong...

[Updated on: Wed, 13 December 2017 19:45]

  •  
matti763

Messages: 17
Karma: 6
Send a private message to this user
Well yes, it could run up to 1000 users. Your issues are most likely with the virtualization. Running VM with 16 vCPU is just wrong. Google "CPU Commitment" and learn the basics. Then the storage, what type of storage this is?
  •  
phcranston

Messages: 21

Karma: 4
Send a private message to this user
I agree with <_at_>matti763. We managed over a dozen different Kerio installs. Our largest install is about 800 users and we don't see any performance issues, but you do have to have the right hardware. Mail servers do a lot of small read/write transactions so having really fast storage is key to good performance. I take SSD RAID when I can get it, especially as the number of accounts goes up. I also prefer running my larger installs on a unix appliance vs Mac OS or Windows OS. In my experience it's a bit more stable and performs better but that could easily just be personal preference. We have several smaller installs on Mac OS that run just fine.
  •  
Bud Durland

Messages: 478
Karma: 59
Send a private message to this user
I have 350+ users on Kerio 9.2, in a VMWare environment. Having done the transition a couple years ago, I can confirm that performance gets much better if you use Linux on the server. We are using Debian, but I don't think the flavor of Linux used makes much difference. I can second the other recommendations:

  • CPUs -- you don't need more than four, and probably 2 is sufficient
  • RAM -- 32gb may be more than needed, but no harm to keep it there
  • DISK -- you definitely want the fastest disks you can afford.
  • FOLDERS -- encourage user to not have more than about 8,000 messages per folder.

Also, the workstation configuration can have an effect on performance. The Kerio offline connector does caching, each user has a copy of their entire message store on their local hard disks. We've found that 8GB of RAM, and a local SSD greatly improves perceived performance.
  •  
Jorge

Messages: 20
Karma: 0
Send a private message to this user
Hi Guys thank you for your answers.
I Read a bit of CPU commitment but still confused about it...

what if I create a new vmware machine with the vmware installation from KERIO's website ( i think its linux base) and give it 6vCPUs and 16gb of RAM and 2TB of storage.
My SAN is a HP LeftHand P4330 24tb raw 24 hp disk running at 7.2K rpms

I don't think I can convince management to deplot a new SSD SAN at the moment...

How would you deploy a server for 350+ users and growing? with tmy current hardware... you input is well appreciated, thank you again!

  •  
bm

Messages: 56
Karma: 11
Send a private message to this user
I'm using Kerio Connect 9.2.5 patch 3 (3336) with 330 users on OS Debian GNU/Linux 8.9, x86_64 in VMWare 6.0 enviroment.

90% of all clients are using Mail.app on MacOS 10.9-.10.11, other clients with OS Window 7 Pro are using Outlook 2010 with Offline Connector on PC Dell OptiPlex 7010 with 240GB SSD and 8GB RAM.

My HW configuration is pretty weak, but everything work like a charm:

CPU: 4vCPU (avg daily usage 5-15%)
RAM: 16GB RAM (avg daily usage 15-30%)
Storage: Local RAID5 (without SSD cache) on Dell PowerEdge R520

[Updated on: Fri, 15 December 2017 18:11]

  •  
Jorge

Messages: 20
Karma: 0
Send a private message to this user
bm wrote on Fri, 15 December 2017 09:02
I'm using Kerio Connect 9.2.5 patch 3 (3336) with 330 users on OS Debian GNU/Linux 8.9, x86_64 in VMWare 6.0 enviroment.

90% of all clients are using Mail.app on MacOS 10.9-.10.11, other clients with OS Window 7 Pro are using Outlook 2010 with Offline Connector on PC Dell OptiPlex 7010 with 240GB SSD and 8GB RAM.

My HW configuration is pretty weak, but everything work like a charm:

CPU: 4vCPU (avg daily usage 5-15%)
RAM: 16GB RAM (avg daily usage 15-30%)
Storage: Local RAID5 (without SSD cache) on Dell PowerEdge R520


Did you use the VMware image? or the installation package for debian to install kerio?
-Thank you for your answer.

  • Attachment: KERIO VER.jpg
    (Size: 69.85KB, Downloaded 350 times)
  •  
Bud Durland

Messages: 478
Karma: 59
Send a private message to this user
Jorge wrote on Thu, 14 December 2017 18:00

what if I create a new vmware machine with the vmware installation from KERIO's website ( i think its linux base) and give it 6vCPUs and 16gb of RAM and 2TB of storage.
My SAN is a HP LeftHand P4330 24tb raw 24 hp disk running at 7.2K rpms


I would install Linux & use the Linux installer for Kerio Connect, rather than the VMWare appliance that Kerio offers. The appliance (which is debian based, last I knew) is a very 'bare bones' version of Linux, and I found that when I wanted to add any tools or features, it was a struggle. If you and your team have limited skill and experience managing Linux, installing a "full" Linux like Debian gives you some GUI tools, and much better choices around the machine's configuration -- disk sizes, network configuration, etc.

The problem with configuring too many CPUs on the VM boils down to something like this: if the VM doesn't actually need the CPUs, the system and the hypervisor (VMWare) spend more time deciding which CPU to use than actually running the system. I run mine on 2 vCPUs, and the CPU utilization rarely goes about 25%.

As an aside, we installed Samba on our to make access to the mail store a little easier. We used that and ROBOCOPY to move the mail store from the Windows server to the new Linux server. Also, be sure to make a separate virtual disk for the mail store. Install Linux in a small (40GB or so) disk, then add a 2TB one to the VM.

Hope this helps
  •  
matti763

Messages: 17
Karma: 6
Send a private message to this user
Jorge wrote on Fri, 15 December 2017 01:00
what if I create a new vmware machine with the vmware installation from KERIO's website ( i think its linux base) and give it 6vCPUs and 16gb of RAM and 2TB of storage. My SAN is a HP LeftHand P4330 24tb raw 24 hp disk running at 7.2K rpms


I would recommend to find out why your current setup is slow or at least, try to measure which part of your infrastructure is the bottleneck. If there is for example some issue with SAN, installing a new VM does not really help.

Can you provide data from Server 2012 Resource Monitor and disk queue from there? If queue is more than 1, then OS cannot write/read disk as fast as it wants.

Can you provide ESXi performance data from storage path? There should be measurements about write latency and read latency.

And can you run test that you shutdown VM, edit ammount of vCPU to 2 and then start VM.
  •  
Jorge

Messages: 20
Karma: 0
Send a private message to this user
Bud Durland wrote on Sat, 16 December 2017 05:28
Jorge wrote on Thu, 14 December 2017 18:00

what if I create a new vmware machine with the vmware installation from KERIO's website ( i think its linux base) and give it 6vCPUs and 16gb of RAM and 2TB of storage.
My SAN is a HP LeftHand P4330 24tb raw 24 hp disk running at 7.2K rpms


I would install Linux & use the Linux installer for Kerio Connect, rather than the VMWare appliance that Kerio offers. The appliance (which is debian based, last I knew) is a very 'bare bones' version of Linux, and I found that when I wanted to add any tools or features, it was a struggle. If you and your team have limited skill and experience managing Linux, installing a "full" Linux like Debian gives you some GUI tools, and much better choices around the machine's configuration -- disk sizes, network configuration, etc.

The problem with configuring too many CPUs on the VM boils down to something like this: if the VM doesn't actually need the CPUs, the system and the hypervisor (VMWare) spend more time deciding which CPU to use than actually running the system. I run mine on 2 vCPUs, and the CPU utilization rarely goes about 25%.

As an aside, we installed Samba on our to make access to the mail store a little easier. We used that and ROBOCOPY to move the mail store from the Windows server to the new Linux server. Also, be sure to make a separate virtual disk for the mail store. Install Linux in a small (40GB or so) disk, then add a 2TB one to the VM.

Hope this helps


Thank you for your time and for your answer I will definitely try all of this.
  •  
Jorge

Messages: 20
Karma: 0
Send a private message to this user
[img]./fa/4775/0/[/img]matti763 wrote on Sun, 17 December 2017 01:12
Jorge wrote on Fri, 15 December 2017 01:00
what if I create a new vmware machine with the vmware installation from KERIO's website ( i think its linux base) and give it 6vCPUs and 16gb of RAM and 2TB of storage. My SAN is a HP LeftHand P4330 24tb raw 24 hp disk running at 7.2K rpms


I would recommend to find out why your current setup is slow or at least, try to measure which part of your infrastructure is the bottleneck. If there is for example some issue with SAN, installing a new VM does not really help.

Can you provide data from Server 2012 Resource Monitor and disk queue from there? If queue is more than 1, then OS cannot write/read disk as fast as it wants.

Can you provide ESXi performance data from storage path? There should be measurements about write latency and read latency.

And can you run test that you shutdown VM, edit ammount of vCPU to 2 and then start VM.


You are right.
From the resource monitors disk queue length for disk D: goes from 8.00 to 20.35
I made the mistake of creating only 1 lun for C: and D: - its a raid5 I think on new server i will use raid10 to speed this up and use full/fat provisioning.
but the disk queue length for C: is always at 0

I added an email of the datastore from vmware.

Thank you for your time and help!


[img]./fa/4775/0/[/img]

  •  
matti763

Messages: 17
Karma: 6
Send a private message to this user
It is quite clear that this is the reason for bad performance. But now you need to figure out why and that could be hard job. I recommended that you drill down to ESXi and check what your host can see.

Use esxtop as this article describe: https://kb.vmware.com/s/article/1008205

Here is good explanation what these values are: https://blogs.vmware.com/vsphere/2012/05/troubleshooting-sto rage-performance-in-vsphere-part-1-the-basics.html

Basically you can see if the issue is related to VM, Host or Storage.

I'm not familiar with HP storage but usually high write latency means that you have write cache disabled or write cache is always full. There could be for example broken cache battery. Check if there is any alerts about that. As you can see read latency, 5ms is quite realistic value when you need to read data from HDD and deliver data to VM. However, 16ms average write latency is not normal.
  •  
Jorge

Messages: 20
Karma: 0
Send a private message to this user
Happy new year Guys!!! and thank you for all your answers and support.
I have one more question hehe if we move to Linux Redhat can I just copy and paste all the Mail storage? are the files the same?

and do you have any other tips on this windows to Linux migration?

Thank you again!
Bud Durland

Messages: 478
Karma: 59
Send a private message to this user
Depending on your mail store size and tolerance for downtime, you might want to use the officially documented method. If not, it is basically a copy/paste, and editing the mailserver.cfg file for OS specific settings. In our Windows to linux move, I mounted the old (windows) mail store via NFS, and used rsync to make an initial copy to the new Linux server (it ran for almost 40 hours). Then I shutdown the windows server, did a 'catch up' copy, and brought up the Linux server. other than having to rebuild a few indexes, it worked well.
Previous Topic: Migration Time from Mac to VM
Next Topic: gmx problem
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: Sun Nov 18 23:20:04 CET 2018

Total time taken to generate the page: 0.78355 seconds
.:: Contact :: Home ::.
Powered by: FUDforum 3.0.4.