Slow Docker updates, extracting slow on new hardware

[ZERØ]

I recently upgraded my motherboard, ram, and CPU in my unraid server. Now it seems like docker updates take a long time to extract.

 

Previous Hardware:

• Gigabyte H55M-S2V Motherboard
• Intel Core i5-660
• 8GB G Skill DDR3 1333

 

New Hardware:

• Supermicro X8SIE-LN4
• Intel Xeon X3470
• 16GB Samsung DDR3 1333 ECC Registered

 

It's the same processor generation and socket (LGA1156). The Xeon is about the same clock rate, but double the cores. I have boost enabled. Any ideas why extracting would take longer? Maybe a bios setting I set incorrectly for the CPU? The processor usage when extracting will use 10-20% of the CPU, jump to 60-80% for a moment, then drop down again. 2-3 of the 8 cores will be around 100% when that happens. For an example, I just updated my home assistant core docker, and the 250 mb piece took 2-3 minutes to extract.

 

On current unraid 6.11.5

Edited January 17, 2023 by ZERØ
Added unraid version

[JorgeB]

You can do a quick CPU test and compare with the old board/CPU if still available to rule that out:

 

simple CPU speed test:

time $(i=0; while (( i < 9999999 )); do (( i ++ )); done)

(this will return the the time required to crunch the integers between 0 to 9999999)

[Squid]

And where is the docker image located?  On the array or on the cache drive / pool?

[ZERØ]

27 minutes ago, Squid said:

And where is the docker image located?  On the array or on the cache drive / pool?

Located on the Cache drive. In the docker settings:

/mnt/user/system/docker/docker.img

Which I can find by navigating from the Main tab and browsing the cache disk. Appdata is also set to cache only

[ZERØ]

9 hours ago, JorgeB said:

You can do a quick CPU test and compare with the old board/CPU if still available to rule that out:

 

simple CPU speed test:

time $(i=0; while (( i < 9999999 )); do (( i ++ )); done)

(this will return the the time required to crunch the integers between 0 to 9999999)

Unfortunately my old motherboard died. I was originally only going to upgrade the processor, but the new Xeon overstressed thee old motherboard and blew a couple CPU power mosfets, apparently a common failure point for that board. I don't believe I can use the i5 in the new board to compare, as it doesn't support registered ECC RAM.

 

The result of that test was 32.6 seconds, which probably doesn't mean anything without comparing to the old processor value. The CPU load on the dashboard showed around 35% usage during the test.

[JorgeB]

15 minutes ago, ZERØ said:

The result of that test was 32.6 seconds, which probably doesn't mean anything without comparing to the old processor

I did the same test before with an X3430 and it took 49.5s, based on the passmark score for both yours looks OK, so l likely not a CPU issue.

[ZERØ]

I've got a thread with this same issue over on reddit and that's led to me discovering I have a high iowait time when the extraction is happening, around 30-50 wa when watching top in the console. I have an SSD as cache, and I've benchmarked it to around 250-300 MB/S write in my old hardware, and watching the write speeds in the Main tab UI I can see it's capable of 150-180 MB/s when copying a file over the network to the server, but the UI display seems choppy; it would show values less than 1 MB/s for a few seconds, then spike to 80-150 MB/s, then drop down again. Is that expected behavior? The same thing happens when I'm extracting, very low write values, then a spike of a fast value, then low again.

docker.cfg cache.cfg SanDisk_SSD_PLUS_240GB_210802A00E89-20230117-1305 cache (sdb).txt

[JorgeB]

Make sure it's being regularly trimmed, you could also test with a different SSD is available