Balancing not working as I expected / hoped with an SSD

[Richard]

Hi,

 

I've been using Drivepool for a couple of months now, but last night I encountered a problem where my SSD ran out of space and Drivepool didn't seem to want to rebalance the pool to free up some space, nor would it fall back to writing pinned folders to a different pool drive so I started receiving disk write errors from my applications. It may be that I am not using this software as intended, but I was certainly hoping it would cater for my scenario.

 

My use case is I have a server with 3 x 4TB archive drives and a 50GB SSD, all members in a Drivepool pool. Everything is set to defaults, except in "File Placement", I have set my 3 "working" folders so they can only be placed on the SSD. I have left "allow files to be placed on other disks if.." selected to cater for the SSD getting full. During normal usage the SSD shouldn't get full as I periodically move files into archive folders as they grow to a certain size. One thing I really like about drivepool is that this "move" operation is instantaneous as it clearly leaves the file on the SSD. I was however counting on it performing a background rebalance later and move these large files to the archive disks (since they are no longer in the "working" folders that are pinned to the SSD).

 

So I think my first question is whether this is an acceptable expectation of Drivepool? Should it be working in the background to shuffle these files around to make sure my SSD continues to have free space?

 

And if for some reason the re-balance isn't working, is it reasonable to expect the "allow files to be placed..." rule to work and not receive disk write errors when the SSD is full?

 

This is what I was seeing:

• The UI repeatedly going through "Measuring" operations.
• When it does occasionally pause the measuring, it shows the SSD as needing rebalancing (e.g. "Unduplicated target for re-balancing (-7.78GB)") but doesn't seem to want to do the re-balance.
• The logs have entries about once a second saying: "[CalculateBalanceRatio] Cannot calculate balance ratio, pool is not measured. (Pool mode: PoolModeNormal)"
• The logs will occasionally stop with the "[CalculateBalanceRatio]" entries and instead have a set of "[Rebalance]" entries saying things like "Calculating optimal slack" and "calculating drive fill ratio", but then concludes with "Balance not critical" (although I think it should be since the drive is full!).

I can reproduce the issue by deliberately filling up the SSD in folders that are not pinned to that drive and then kicking off measurement. I've attached a log file extract and some screenshots from a reproduction run. I can manually fix the problem by moving some files from the SSD to another pool member.

 

I have once seen DrivePool perform the re-balance itself, although I don't know if I did something to cause it or if it just suddenly started working. I can tell you that the [FileMover] wrote about 58K log entries for "cannot be moved due to pattern limit" as it listed all the files in my pinned folders about 175 times (so still not a completely smooth operation).

 

Could this be a bug or do I need some more specific configuration to support my use case? I've seen another thread mentioning the "Cannot calculate balance ratio" error but it otherwise didn't seem like a similar problem.

 

Many thanks in advance with any help with this!

 

Richard

 

DrivePool.Service-2015-05-29.log.zip

[Christopher (Drashna)]

Richard,

 

What are the settings for the Balancer settings?

 

Namely, when it balances, and the ratio for which it balances.