Zentyal Server Maintaining Unnecessarily High Load Constantly

[Lonniebiz]

I have a Zentyal server maintaining high load (averaging 2.0 all the time). CPU% is not the problem; that remains relatively normal.

It seems to be an IO issue with the command jbd2-dm-0-8 constantly journaling the file system. This has been going constant for about two weeks, and the jbd2-dm-0-8 command never seems to rest.

iotop :
http://neartalk.com/ss/2013-09-02_18:26:13.png

iotop -obtqqq | grep jbd2 :
http://neartalk.com/ss/2013-09-02_18:35:09.png

Zentyal > Maintenance > Monitoring:
(see attachment)

Background:

I've installed Zentyal 3.0.26 onto a Hyper-V virtual machine to function as a Jabber server for a company that wants inter-office instant messaging. In order to setup user accounts, I've added this Zentyal server as an additional domain controller on an Windows Active Directory domain. Clients are able to consume the service using Pidgin, and by logging in with their Windows domain credentials. However, unlike all the other Zentyal installs I've done (with and without hyper-v), this server maintains an unnecessarily high load, which is strange because it is only designated to perform rather minimal tasks (those necessary to run the jabber service).

[Lonniebiz]

I've attached a diagram representing my system load over the last week. With the exception of reboots, it has remain around 2.0. Any suggestions?

[half_life]

Since I don't use Hyper-V I am not sure of the nomenclature but if I was seeing the VM performing like this under Proxmox I would be questioning why there is a bottleneck writing back to the physical disk.  Look closer at the byte values of the disk writes shown by iotop.  The values are measured in K not M and aren't large at all.  This isn't an issue with the Zentyal setup but an issue with the hypervisor or the physical disks that the VM is stored on.  I would imagine that the iowait percentage shown in top is rather high as well.

Could you tell us more about your setup? 

As a comparison my systems have a 3TB Raid5 array composed of 4 disks on a real raid controller.  Disk reads measure out at around 500MB/s average with bursts half again that high.  Writes are around 260MB/s  with short bursts into the mid 300s (these figures are from memory and are approximate).   Typical system load with about 20 VMs with a few heavy hitters involved is about 5MB/s writes.  The only time my VMs feel any wait related to disk I/O is during backups at night and is related to me having not tuned how dirty_writes are managed.

[Lonniebiz]

The load is still higher than it should be, but I'm here to report that upgrading that virtual machine to Zentyal 3.2.2 has cut its load in half. See the attached screen-shot. You can clearly see when the upgrade to Zentyal 3.2 occurred.