Slow display in “Storage” tab

[mpfusion]

The “Storage” tab feels very sluggish to display its content. Here are some numbers:

Code:

Storage → Active Volumes
15.4 seconds
 
Storage → ZFS Snapshots (10 displayed by default)
1:53 minutes

To compare with the terminal:

Code:

zfs list -t snapshot -r tank
11 seconds 
 
zfs list
0.01 seconds

I understand that the GUI is not as fast as the terminal. But displaying the datasets taking more than ten seconds seems unreasonably slow, let alone almost two minutes to display ten snapshots.

Is this normal performance or is something wrong with my system?

System:
FreeNAS-9.2.1.3-RELEASE-x64
45 data sets
6200 snapshots

G2020
32GiB ECC
two 2x3TB mirrors

 

[cyberjock]

No, I'd say something is a bit odd. Mine normally takes 3-10 seconds depending on how busy the server is. I consider that reasonable.

I'd connect via SSH to the box and run # top. Then check out the same windows in the GUI. I'm betting your abnormally large number of snapshots is a bit much for it to process and python is maxing out a core trying to get it all done. I've seen people have as much as 200-300 snapshots. Never have I seen someone with 6200 snapshots! I could be mistaken but I think the processing of the output is handled as a single thread by the nature of the task, so a large number of snapshots could really hurt performance.

Just a note in case you try trimming down lots of snapshots at once, there's a possibility your pool may go unresponsive due to the I/O that may be necessary. So don't try to do a snapshot remove command that involves an asterisk or anything. It might not go well. And no matter what you do, do NOT reboot the server if it goes unresponsive. You could lose your pool.

 

[mpfusion]

No, I'd say something is a bit odd. Mine normally takes 3-10 seconds depending on how busy the server is. I consider that reasonable.

You're talking about the “ZFS Snapshot” tab which takes 1:53 on my system? The system was idle when I ran the measurements.

I'd connect via SSH to the box and run # top. Then check out the same windows in the GUI. I'm betting your abnormally large number of snapshots is a bit much for it to process and python is maxing out a core trying to get it all done.

The CPU doens't really seem to be that busy:

Code:

8% zfs
3% python2.7
1% ssh

I've seen people have as much as 200-300 snapshots. Never have I seen someone with 6200 snapshots! I could be mistaken but I think the processing of the output is handled as a single thread by the nature of the task, so a large number of snapshots could really hurt performance.

Since FreeNAS was designed for big enterprise storage, I'm surprised to see it choke on a few thousand snapshots and 45 datasets.

For the record, I ran the test on the backup system as well (completely idle, no users, no services except SSH) and the performance was the same. So it might possibly be related to how the “zfs list” parsing is implemented in python. Is it worth filing a bug report?

You seem to wonder why we have so many snapshots. When I set up the system I had two choices. Either a very simple “Periodic Snapshot Tasks” list using recursive snapshots. This keeps it simple but creates a large number of snapshots. The other possibility would have been to implement an “rsnapshot” like schedule. Store 3-hourly snapshots for a day, store daily snapshots for a week, store weekly snapshots for a month, etc. Since FreeNAS didn't support it directly and one had to set it up manually, I went for the first solution. I didn't really see any drawbacks of having more snapshots than strictly required.

 

[mpfusion]

And why does Storage → Active Volumes take 15 seconds to display? It's not supposed to parse the snapshots. And “zfs list” only takes 0.01 seconds. This seems odd. It's just 45 lines to parse.

 

[cyberjock]

For the snapshotting problem, I'd put a bug ticket in. Explain your situation, maybe link this in your ticket. It doesn't appear to be a CPU issue, so the limitation should probably try to be found and torn down wherever that is. I can't guarantee when or if there will be a fix. It could still be CPU bound in some weird obscure way, but I think that's unlikely.

Storage -> Volumes takes me like 3-5 seconds. I have 2 pools and about 15 datasets. I'm wondering if there is something weird/broken with your setup with regard to rendering webpages.

Just for giggles, have you tried a different browser? usually we recommend Firefox around here. Chrome usually works fine too. Usually web problems are because of addons, custom settings, or domain policies that cause problems. IE is just bad.. very bad.

 

[mpfusion]

For the snapshotting problem, I'd put a bug ticket in. Explain your situation, maybe link this in your ticket.

I will do that.

Storage -> Volumes takes me like 3-5 seconds. I have 2 pools and about 15 datasets. I'm wondering if there is something weird/broken with your setup with regard to rendering webpages.

Just for giggles, have you tried a different browser? usually we recommend Firefox around here. Chrome usually works fine too. Usually web problems are because of addons, custom settings, or domain policies that cause problems. IE is just bad.. very bad.

I tried Chromium and Firefox, both behave the same. I think it's a FreeNAS issue and unrelated to the rendering.

If 15 datasets take five seconds on your machine and 45 take around fifteen seconds on mine it's both around three datasets per second, which is darn slow in my opinion. There's definitely some very inefficient processing being done.

 

[eraser]

To link this thread to the bug ticket for future readers: https://bugs.freenas.org/issues/4737

 

[Jeroen Verburg]

Is there a solutions for this problem? I see that the issue is old.
I have 2 freenas with ZFS replication. (9.2.1.7) each with 8 gb off ram
The send the replication to each other. Both machine have a slow respons on the snapshot tab.
The first time I open the active volumes it is slow. The next time it respons ok (in the same firefox browser session)

In the shell; Active volumes
the volumes are listed very rapidly. (zfs list) (each machine have 3 volumes )
In the shell; Snapshots
the zfs list -t snapshot option is fast on one machine and slow (about 2 -3 minutes) in the other machine.

Can there be an issue with the replicas to each server?

 

[Jeroen Verburg]

Never mind my problem, when I created the replication task I did not vink ,
Recursively replicate and remove stale snapshot on remote side:
in the replication window.
There fore I had 7400 remote snapshots from the other side.
I am removing them by hand and hoping the new ones will go automaticly