Initializing File-IO Volumes

[Laxity]

A possibility to set the priority for the init. task would be nice.

I am right now initializing a 1.4TB Volume.

The VMs that are running on the XenServer hosts are terribly slow now because of this.
I would not mind if the init. procedure would take twice as long or even longer if my VMs would work normally then..

It has nothing to do with CPU or RAM in my case. RAM usage is at 7% and CPUs are at normal temperature. BTW: is there a way to look at the CPU usage using the console? I can only see my CPU Temperatures using the IMPI module..

The init is quite fast -aprox. 1.5h for an 1.4TB volume on a Raid-6 (5x1TB discs on areca 1261 controller).

[Sh-J]

Dear Laxity,

THe CPU tempreture can be only seen from your IMPI card. But you can get other server stats from the check-Sys option in the system GUI. (STATUS_>Hardware)

Regards,
SJ

[Laxity]

Thanks Sh-J,

I discovered the link to the statistics with your help

[Laxity]

But the tool does not store history data as it seems.. When I click the button to display the values it always only gives me back the values of the last 15 minutes or so:



do I have to enable history logging somewhere?

Thanks in advance
Laxity

[Laxity]

tried it again right now.. again only the data for the last 2 minutes is displayed..

any hints would be great! Thanks!

[Sh-J]

Dear Laxity,

I tested it on my server and I can see stats for long periods.
What release you are using?

Regards,
SJ

[Laxity]

Hi Sh-J,

I am using Version: 5.0.DB49000000.3278

Regards,
Laxity

[jmo]

A possibility to set the priority for the init. task would be nice.

I am right now initializing a 1.4TB Volume.

The VMs that are running on the XenServer hosts are terribly slow now because of this.
I would not mind if the init. procedure would take twice as long or even longer if my VMs would work normally then..

It has nothing to do with CPU or RAM in my case. RAM usage is at 7% and CPUs are at normal temperature. BTW: is there a way to look at the CPU usage using the console? I can only see my CPU Temperatures using the IMPI module..

The init is quite fast -aprox. 1.5h for an 1.4TB volume on a Raid-6 (5x1TB discs on areca 1261 controller).

A happy new year to you all!

And now to something complete different...

I've run an init of a 100GB FC volume yesterday... it took only a good handful of minutes, but caused a load of 26... all FC targets for our Xen server disconnected because of time-outs. I had quite a recovery job to do, a Sunday favorite of mine :-(

(We as well had no major CPU usage, memory was mostly free (we have 4GB in the machine, running in 32 bit mode with mostly FC targets that sure is a waste of memory), our RAID peaked at 190k sectors/s writes, which seems close to the maximum to be expected) And please don't blame the disks... as the disks are always the bottleneck, it's only a question of the FC volume size to bring the system to its limits with even the fastest disks. A proper system design would limit the priority of the init job to allow the productive resources to remain usable.

This is definitely a show-stopper for proper production use, we'll have to shut down all VMs prior to creating even mid-size (100GB nd above) FC targets :-(

With regards,
Jens

[enealDC]

If i were you, I just would not bother to use file-io volumes. Stick with block-mode and suffer the performance penalty. This was my decision because I was trying to avoid this very same issue.

It's not unusual to make adds/remove to a SAN during business hours. The action should not cause a major service interruption..

[To-M]

JMO,

Good point that you make and a topic that we should add to our knowledgebase.

I agree that this will cause a performance hit due to the fact that it is using the dd command for the initialization. The best practice would be if you have a new system with 12TB (or whatever size) create the full size of 12TB (FC or iSCSI) Logical Volume and set it for the initialization. Once complete then you can delete it and you can create your logical volumes without. This will be the best way and know that your volume is good.

Other way would be to use the Volume Replication and failover to the destination server making it Source. Then once the init is completed failback, I know both can be time consuming vs the performance issue.