KVM call trace & eth0/eth1 drop

[therapist]

Running 6.7.0-rc6 on a supermicro x9dri-ln4f w/ 2x E5-2670v2 & 128gb ECC ram

1x 9211-8i -- 6x drives on array (slot 4)

1x 9207-8i -- 5x non array SSDs + 1x cache SSD (slot 6)

1x 9207-8e -- 4x array drives + 2x non array drives (slot 1)

gigabyte gtx 1070 (slot 3)

11 array drives -- 6tb parity

 

Upgraded from x8dti-f w/ 2x l6540 in late January / early February -- had no issues with this system other than lacking a little grunt for PLEX transcoding

 

Ive been having recent issues with call traces which seem to be occurring during parity checks.  Noticed them during the previous month's scheduled partiy check (2/28)

I had my tunables set to:

Tunable (nr_requests): 128

Tunable (md_num_stripes): 4096

Tunable (md_sync_window): 2048

Tunable (md_sync_thresh): 2000

And I've lowered sync_thresh to 192 which alleviated a number of the call traces I was seeing but don't understand why this became an issue after the upgrade when I re-used all hardware other than the mobo/cpu

 

Also, still getting a KVM related call trace at the onset of the parity check -- and NIC dropouts while parity check is running

 

Easy answer is to do parity checks less (I'm on UPS & SONNEN battery) but I'd like to fix the problem...

Any advice on where to go?

ocho-diagnostics-20190401-1838.zip