Very Slow Parity Sync for New 32TB 4+1 Drive Array

[SLRist]

Hi - just built a brand new unRAID server in a HP Proliant N40L 2GB server chassis (I already have 4 other unRAID servers in similar hardware - except they have 6TB drives or smaller) 4 data drives + 1 parity. These are shucked 8TB Seagate desktop 5400rpm drives which have been extensively tested for errors (over several days) before installing. I also ran a memory test of the 2GB single memory card on the motherboard before starting.

 

Running initial parity sync on the empty array. It's running *very* slowly (1MB - 6MB per second) - sync is only at 6.2% after 34 hours. Estimate at this rate is another 20 days to complete!

 

I have attached a Diagnostics export. I see lots of repeated mention of:

 

Dec  7 02:17:20 UNRAID-05 nginx: 2022/12/07 02:17:20 [error] 1353#1353: *390296 nchan: error publishing message (HTTP status code 500), client: unix:, server: , request: "POST /pub/devices?buffer_length=1 HTTP/1.1", host: "localhost"
Dec  7 02:17:20 UNRAID-05 nginx: 2022/12/07 02:17:20 [error] 1353#1353: MEMSTORE:00: can't create shared message for channel /devices
Dec  7 02:17:21 UNRAID-05 nginx: 2022/12/07 02:17:21 [crit] 1353#1353: ngx_slab_alloc() failed: no memory
Dec  7 02:17:21 UNRAID-05 nginx: 2022/12/07 02:17:21 [error] 1353#1353: shpool alloc failed
Dec  7 02:17:21 UNRAID-05 nginx: 2022/12/07 02:17:21 [error] 1353#1353: nchan: Out of shared memory while allocating message of size 7993. Increase nchan_max_reserved_memory.

 

Any suggestions please?

unraid-05-diagnostics-20221208-0759.zip

[Kilrah]

Minimum RAM for current Unraid is 4GB.

 

Also it seems these drives are SMR, usually Seagate ones behave decently in sequential writes but that could also be a cause.

Edited December 8, 2022 by Kilrah

[JorgeB]

Also make sure write cache is enabled in the server BIOS, it's disabled by default.

[SLRist]

Thanks for your suggestions. Is it OK for me to pause the parity sync, shut down the server and upgrade the RAM before re-starting and continuing the parity sync, or would I be better to scrub the array and start from scratch?

Edited December 8, 2022 by SLRist

[itimpi]

8 minutes ago, SLRist said:

Thanks for your suggestions. Is it OK for me to pause the parity sync, shut down the server and upgrade the RAM before re-starting and continuing the parity sync, or would I be better to scrub the array and start from scratch?

By default Unraid will restart the parity sync from the beginning.   If you have the Parity Check Tuning plugin installed then it will attempt to restart the parity sync from the point reached (assuming you have the restart option set in the plugin settings) when you next start the array.

[SLRist]

Many thanks. To be honest, I think I'll just let it start from the beginning. Hopefully it will run a lost faster after the RAM is upgraded and maybe write cache is BIOS enabled (if it isn't).

[SLRist]

OK - I temporarily borrowed 8GB of RAM from one of my other servers and enabled write caching in BIOS - now parity sync speed has gone from about 4MB/s to 100MB/s!

 

Estimated completion of sync is 22 hours rather than closer to 22 days!

 

Thanks for the help. Problem (apparently) solved!

 

Is it a good idea to leave write cache enabled permanently - or is it better to disable it?  All my servers are connected to a UPS.

Edited December 8, 2022 by SLRist

[JorgeB]

15 minutes ago, SLRist said:

Is it a good idea to leave write cache enabled permanently

I recommend this, especially since you have a UPS, or write performance will suffer a lot.

[SLRist]

Perfect - thanks. Maybe I should enable it on all my 4 other unRAID servers too