High I/O wait -- Do we need to upgrade?

sclemons

New Member
We've been experiencing intermittently high I/O waits for some time on our server. Roughly once a week, I/O load will be high for a 12-18 hour period before returning to normal. The waits have a direct impact on the web server performance, mainly by slowing db writes.

At this point, I've been through several rounds of optimization with the software vendor, and essentially exhausted that resource. They have a full breakdown of my hardware/VPS specs, and they say there's no reason I should have any resource problems.

I've had several rounds of support tickets with KH techs, and I can honestly say that they've been more trouble than they were worth. I won't rehash those issues here, but will say that I think KH should be more careful about letting first-tier support techs with limited bash knowledge log into customer machines as root. Additionally, KH needs a better system of ticket escalation.

That said, I'm still having problems and hopefully have a few simple questions: is there a problem with the server hosting my VPS? Is there a combination of VPS clients on that machine (mine included) that occasionally exceed the disk I/O capabilities of the host machine?

If the host machine checks out, would I benefit instead from upgrading to a hybrid VPS?
 
sclemons,

I apologize for the delay in getting back to you on this. Forum is not the official communications channel and as such is not being monitored all the time.

I would appreciate if you can please PM me with the ticket number where your situation is described and where you did not receive appropriate help from our team. I'll be happy to review the situation for you.

Seeing high I/O wait inside the Virtuzzo v4 VPS means that there is enough disk i/o activity inside the VPS to trigger i/o calls queuing and, as such, produce (noticeable) i/o wait timing. In majority of cases high i/o wait problems needs to be and can be resolved inside the VPS while in some cases external factors such as raid-level problems, filesystem problems, i/o queue saturation, etc may cause i/o problems. When it comes to "external" problems we do monitor machines for such things and do take proactive action if any of the problems detected. When it comes to in-VPS issues - this highly depends on the processes that produce most of the i/o calls, what kind of i/o requests are being generated and so on.
As for the upgrade to hybrid - it might or might not help. From one side Hybrid servers house smaller number of accounts, as such general availability of CPU power and disk I/O might be a bit better, on the other side moving VPS filesystem to a Hybrid server will help to get rid of possible fragmentation which could be a limiting factor when it comes so things like selects from large database tables, read/writes from/to directories with significant number of files (not a good idea regardless of how you look at it) and so on.

Regards,
Paul
 
Thanks Paul. You have a PM.

Actually, I take that back. I don't have permission to send you a PM.

The ticket numbers are 119001 and 113372
 
sclemons,

Thank you. I quickly scanned both tickets and could see that in the older ticket our team was able to catch the moment when issue happened and pointed out why this is happening despite the fact that you're having a self-managed VPS with no control panel installed. In the second ticket I see that we weren't able to catch the moment when problem exists as such no pointers were provided however taking in account that you have a self-managed VPS it isn't really our job to watch things inside the VPS in attempt to catch situations like we were able to catch in your first ticket.
I'm going to follow up with you in your most recent ticket with my initial observations. And, yes, based on what I'm seeing so far your experience with Hybrid should be better than with just regular VPS.

Regards,
Paul
 
Top