Writing to cache: degraded speed

[stomp]

Hello,

 

I recently changed my cache drive, went from SATA SSD to M.2 NVMe. The goal was to enjoy better writing speed on my 10 GbE network. At first everything went fine, speed increased from around 500 MB/s to 750 MB/s and was very stable. At that point I thought my CPU was the bottleneck with high CPU usage during writes (90% on average).

 

Now for a couple weeks, writing speed has drastically degraded. It usually starts at 750 MB/s but about 30 sec later goes to around 120 MB/s and stays that way. Sometimes it even stops to write, then goes back on. CPU usage is unaffected, staying close to 90%. I cannot remember having done any software or hardware change in the meantime.

tower-diagnostics-20190128-2212.zip

Log looks okay to me and I don't know how to monitor for anything that could slow down the process. I usually hear something in the server when the speed goes down, probably some HDD doing some things, can't say what's happening though.

 

I already tried to restart the server, disable docker and uninstall every plugin.

 

You'll find the diagnostic attached.

 

Thanks in advance.

 

EDIT: I also ran DiskSpeed to check the drive speed per se and it varies between 750 and 1000 MB/s on the all drive (512 GB).

[JorgeB]

Install the Dynamix trim plugin and scheduled it to run daily, it should make a difference.

[stomp]

Thanks.

I ran TRIM and just did one test. It seems like it fixed my problem.

I thought that by 2019 TRIM was standard in everything, seems like I was wrong :-)