Slow Backup to 8Tb Seagate drive, killing the NAS performance

[tekguru]

The backup complete overnight. CPU load this morning looking like:

But IO Wait is still peaking occasionally at 21 to 47%. Is that within normal expectations?

I'm rebooting the NAS just in case.

 

[Rusty]

The backup complete overnight. CPU load this morning looking like:
View attachment 11099

But IO Wait is still peaking occasionally at 21 to 47%. Is that within normal expectations?

I'm rebooting the NAS just in case.

Guessing the volume chart is also peaking?

 

[tekguru]

To help clear out any rogue processes I've rebooted the NAS, so not a lot of stats data collected yet, but the forum is positively zooming along and stats are starting to look a lot better. The volume chart is looking okay too I think?

I'll post more stats later when further data has been collected.

 

[Rusty]

The volume chart is looking okay too I think?

Now it looks like all is well. Guess its time to get yet another box Guru! Share the tasks across multiple units

 

[tekguru]

Yeah, the CFO (wife) would definitely not approve spending on a second NAS

One thing I noticed after things looked good, when a Tmemachine backup started on my Mac that the I/O wait ramped up and so did the other meters. Once the backup completed things did not decrease again.... So I'm wondering if I need to change settings on the drives that are used for TimeMachine backups?
If you look at the images below you should be able to see what TimeMachine kicked in.

 

[Rusty]

That means that the in background the volume is doing some housekeeping process. This is depend on the number of versions, size of the volume (free vs used) that might expain why this has started in the 1st place or happens after a "major" task like a backup.

If the TM backup destination is under quota, and the space is getting smaller, TM will start to FIFO rotate the data. That will ofc, cause again spikes that migh continue after the actual backup process has completed.

 

[tekguru]

Understood. The used figure on the share is "864.36 GB containing 27672 files, the quota is set at 1536 Gb. The HD being backed up is my main drive and is capacity 499.96 Gb, used 350.87Gb.
Settings look okay:

And the Time Machine folders are set properly.

The CPU load has reduced but still higher than it was prior to the TM backup starting.

Or is this still considered within normal operational bounds?

 

[Rusty]

Or is this still considered within normal operational bounds?

Hard to say without continuous metrics being compared.

 

[tekguru]

Agreed, everything seems to be running smoothly. Looking back at the CPU usage:

The stress from the USB drive backup can be clearly seen (16th to 19th). It looks like there has been high I/O wait in the past, but likely when TM was working. I've now used TimeMachineSchedular to reduce the TM backup frequency and we'll see how it goes.
At the moment IOW is sitting at round 28-30% and there are no issues with the forum (which was my main concern). I'm holding off re-enabling Docker / Homebridege for the moment though as I know they tend to raise CPU levels by 8-10%.

 

[Rusty]

The stress from the USB drive backup can be clearly seen (16th to 19th).

Yep that clearly states that the 16th/17th onwards was the date when things started to get problematic.

It could also mean that some of the drives is starting to get problematic, even if there are no smart errors. If this continues like this with all the services that were active before and atm are not, that could also mean a volume/drive(s) issue.

 

[tekguru]

Heck I'll continue to monitor then I guess. The NAS was only ordered on the 28 March 2020, and was ordered with 4 x 3TB Seagate IronWolf Drives, so I'd expect a decent lifespan? The volumes and drives are not showing any issues at all.