Help diagnosing cause of pauses during 10Gb cifs transfers

[xnaron]

I'm setting up a new freenas box to replace my old one. It's in the testing phase right now. Basically I am having long occasional pausing during a 10G cifs transfer. This has occured during windows 10 and ubuntu 18.04 transfers. To get a feel for what is happening please watch this video. I have zpool iostat 1 running and also the web status gui of my 10Gb switch.

https://youtu.be/tCsfPhiVho4

- test file is 42GB generated with dd if=/dev/random...
- Have tried both chelsio and intel 10Gb network cards in the freenas box with same results.
- sometimes the copy completes without pausing
- have seen as many as 2 pauses and sometimes only 1 pause
- The error screen on the 10Gb switch shows no errors for any of the ports
- no other load on the freenas server as it is not in production and this is the only zpool.


System:
ibm x3650 m3, 2x X5650 2.67Ghz cpu, 128GB ram
Netapp DS4246 enclosures ( 2 of them) IOM6
4TB Seagate Ironwolf x 8 in Raidz2 pool. 4 disks in each ds4246
Netapp controller - PMC-Sierra PM8003 connected with 2 cables one to each ds4246
Freenas 11.1 U6

I've got an LSI 9200-8e flashed to IT mode waiting to try as soon as the 2 qsfp -> mini sas cables arrive on Tuesday

Looking for advice on things I can monitor during the transfer to help narrow down the issue.

Thanks

 

[kdragon75]

Take a look at zpool iostat vol1test -v to see if you have any issues with a particular drive.

 

[HoneyBadger]

Take a look at zpool iostat vol1test -v to see if you have any issues with a particular drive.

SMART might also give some hints but this feels like a bad drive/cable - it's not the usual "async writes clobbering your pool" because the iostat is showing your pool itself come to a halt, not just the network/ingest into the transaction group.