Parity sync - individual disk reads drastically different?

[tyrindor]

I've been having a run of bad luck lately. I had to downgrade to B14 because the the 5.0 RCs are having issues with SASMV8 cards. However, now that I downgraded to B14, i'm noticing weird behavior that was not happening on the RCs. Disk 1 is showing incredibly lower reads during parity sync. I'm not sure if this is normal or not. I show 0 SMART errors on all drives, no syslog messages after parity sync was started, and 0 sync errors. These are all connected to the same 3 SAS2LP cards, so I assume reads should be similar (because they were on RC5). Stats were cleared before running parity sync. Server has not been accessed.

 

Should I be concerned? I'm not 100% sure on how this process works, so I'd rather be safe than sorry. I planned on upgrading Disk1 to 3TB this week.

 

Thanks!

[tyrindor]

Here's the reads when parity is at 1.91TB of 3TB, the gap is even larger now. The 2TB drives (including disk1) will stop being read from soon. How is it possibly correctly calculating parity with such a small amount of reads from Disk1? The other 2TB drives of the same model, on the same SAS cards have almost 4x more reads.

 

Maybe i'm worrying about nothing but i've never seen this before. =/

 

[aatti]

Did you ever receive any feedback from anyone? I've just noticed very similar behaviour .... on disk 1 all 4 disks are 3TB / unraid 6.

[bonienl]

Did you ever receive any feedback from anyone? I've just noticed very similar behaviour .... on disk 1 all 4 disks are 3TB / unraid 6.

 

Wow, you ask for a response to a question raised more than 3 years ago ...

[RobJ]

Those read and write counts have never made complete sense, as they seem to come from a low level, where block sizes are affected by a number of things, including the drive models, disk controller, sector size (512 vs 4K), caching and buffering, and other factors.  Usually, the numbers aren't as far off as the above, but they can be.  No one has ever found anything wrong associated with the disparities.

[aatti]

I know ... old topic but actually never had any issue with the system and I could not find anything wrong. Never payed attention to this detail (was running unraid 4, 5 with 1.5T disks in the past) and probably the stats were not clean when parity check was started. Anyway, I can post a picture tomorrow as I'll run a new parity check (counters reset) and probably I'll get the same, huge difference even if all disks are 3T / 4K aligned, xfs and two of them are almost identically filled and the same model (WD Red 30efrx) but different firmware.

 

I'll post a picture as promised, but won't dig further this night.

Thanks anyway

 

 

[aatti]

As promised, here it is.

Any idea is welcomed, otherwise I may drop a line to Tom.

 

What I could do eventually... is to stop the system, remove the drive and copy the full content - just don't have enough space on external drives. But I'm almost certain there is no problem with the content and I did not find anything wrong in logs.

[JorgeB]

Don’t know the reason but this is perfectly normal, see my test server below, parity check after turning server on, same model SSD have same firmware, all SSDs on same controller and 4k aligned.

 

[bonienl]

I like your parity check time, just 11 minutes 

[JorgeB]

Yeah, but should be a lot faster, there’s an issue with v6 and the SAS2LP, that’s why I use this server to test the latest betas.