Delays every 12 mins!

[akaVintz]

Hi all,

I was wondering if anyone could give me any advice on how and where I should start trying to diagnose my problem?

I have created a new zfs volume 4x1TB drives in RaidZ2, not ideal but im working with what I have.

The thing is I am copying a large 630GB file to the vmware datastore I have sat on this lun.

I thought the issue may be with the win 7 laptop I am copying from but the reporting on freenas shows a lack of network activity and cpu usage every, exactly, 12 mins. During this time my copy drops to 1MBps and the refresh on the datastore is about 60seconds. After this 12 min lul I get 12 mins of 26MBps and the refresh is instantaneous. The laptop is connected via copper to a 1GB switch, I am running vmware esx4 and the datastore is a vmfs3 with block size of 3MB. Im using iSCSI.

Any advice would be greatly received.

Merry xmas

V

 

[dlavigne]

Which FreeNAS version?

 

[akaVintz]

Oops its 8.3.1.

 

[diedrichg]

How much RAM?

 

[akaVintz]

4GB

 

[Ericloewe]

6GB was the old minimum, but 8GB was safer. 8GB is the minimum for ZFS on newer versions.

 

[akaVintz]

Thanks guys, ill increase the ram and monitor.

Merry xmas.

 

[L]

The candence is extremely odd. 12 minutes..I would also upgrade your version. Hard to hunt down bugs that may have been fixed a long time ago.

 

[diedrichg]

I was just thinking that because it was so perfectly timed that the RAM was being filled. During the lull FN chewed through the data and then resumed filling the RAM again. Just a theory...

 

[L]

That flush is done every 5 seconds on current deployments. Not 12 minutes. Even in earlier zfs the flush timeout was 30 seconds. It is really strange. I saw one like this recently where there was another router on the network and the system would take another longer path.

I looked though a ton of the config inside the freenas to just see if i could find a 12 anywhere. There was one cron which does save-entropy which is at the 11th minute of every hour. Still not sure what that does..

 

[zambanini]

/var/log/messages output may help, otherwise do a endless iperf test with a script. not sure if iperf was available in freenas 8.3 ;)

you could also test if the problem only exists with iscsi...create a nfs datastore.

 

[akaVintz]

Thanks everyone, this is really useful. Ill get the RAM up to at least the min recomended and then see how it goes, I will also try the nfs store.