FN 11.1-U1 appears to solve the scrub problem of 11.1

[danb35]

Some folks, including me, had problems with 11.1-RELEASE and scrubs, resulting in nasty things like system lockups and/or scrubs that would slow to 0. A number of us reverted to 11.0 for that reason. iX said that 11.1-U1 was supposed to fix the relevant bug, and it looks like it did.

I took the plunge and updated to 11.1-U1 last night. This issue does seem to be resolved, and scrubs much faster than under 11.0. For comparison, the last scrub of my pool took 49 hours. After the update, I started another scrub. It completed in 19:45, 60% less time than under 11.0, under usage as similar as I could make it. Sounds like a winner.

 

[sef]

Yay! Now we are so happy we do the dance of joy!

 

[INCSlayer]

the media embed seems to be "slightly" larger than i expected

 

[Brett Davis]

Some folks, including me, had problems with 11.1-RELEASE and scrubs, resulting in nasty things like system lockups and/or scrubs that would slow to 0. A number of us reverted to 11.0 for that reason. iX said that 11.1-U1 was supposed to fix the relevant bug, and it looks like it did.

I took the plunge and updated to 11.1-U1 last night. This issue does seem to be resolved, and scrubs much faster than under 11.0. For comparison, the last scrub of my pool took 49 hours. After the update, I started another scrub. It completed in 19:45, 60% less time than under 11.0, under usage as similar as I could make it. Sounds like a winner.

What's the size of the pool?

 

[danb35]

What's the size of the pool?

Specs are in my .sig. 48 TB net capacity, around 30 TB used.

 

[Brett Davis]

Specs are in my .sig. 48 TB net capacity, around 30 TB used.

Ah, sorry, didn't think to look there. That's impressive for a pool that size!

 

[ethereal]

this is surprising to me i thought the scrub speed was related to how much data you have and how fast your discs are. years ago my pool which was generally the same size each month and took 19 hours to scrub went to over 50 hours - this was one of the reasons i went from fortnightly to monthly scrubs. it was this month i discovered that the scrubs are going faster.

i didn't think that different versions of freenas were maybe affecting it

 

[danb35]

i didn't think that different versions of freenas were maybe affecting it

I don't know that there have been major changes to the code over time, but there were with the release of FreeBSD 11.1--though I don't readily find a link describing them.

 

[ethereal]

pool Storage 87% full - scrub time 10:43:52 was my latest scrub

my pool used to take 19 hours - then 72 hours and now 11 hours (my pool is fuller now but scrubs much faster)

 

[Scharbag]

this is surprising to me i thought the scrub speed was related to how much data you have and how fast your discs are. years ago my pool which was generally the same size each month and took 19 hours to scrub went to over 50 hours - this was one of the reasons i went from fortnightly to monthly scrubs. it was this month i discovered that the scrubs are going faster.

i didn't think that different versions of freenas were maybe affecting it

I cannot remember where the link is but there is a video describing why scrubs are faster now. It has to do with using memory to intelligently prioritize the scrub. This was introduced to FreeNAS in 11.1.

My pools scrub way faster now too.

:)

 

[rs225]

When it got slower, it is probably because of changes to the default priority given to scrub I/O. This minimized the effect of scrubs on normal operation.

When it got faster, that was the new sequential resilver/scrub code. That code reads a large chunk of metadata into memory, determines all the blocks it references, sorts it, and then reads and verifies it in disk order. When it runs out, it repeats. The old method read blocks in order of when they were written, generally. It was reasonably in order, considering no effort required.

 

[danb35]

that was the new sequential resilver/scrub code.

I was wondering if it would affect resilvers as well--sounds like it would. That's also good news.