[Bug 235856] FreeBSD freezes on AWS EC2 t3 machines

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235856

--- Comment #18 from mail at rubenvos.com ---
Hi Colin,

> 1. How repeatable is this?  Does it happen to every instance you launch (after a variable number of days)?

Unfortunately we are not in the habit of often redeplying our zfs nodes (since
they provide storage for whole platforms) :( It still happens to the affected
nodes though:

 bzgrep "nvme1: Missing interrupt" /var/log/messages.0.bz2 ; grep  "nvme1:
Missing interrupt" /var/log/messages
Nov 16 03:04:18 zfs01 kernel: nvme1: Missing interrupt
Nov 16 03:05:19 zfs01 kernel: nvme1: Missing interrupt
Nov 25 03:04:36 zfs01 kernel: nvme1: Missing interrupt
Nov 25 03:05:07 zfs01 kernel: nvme1: Missing interrupt
Nov 25 03:06:07 zfs01 kernel: nvme1: Missing interrupt
Dec 13 03:04:34 zfs01 kernel: nvme1: Missing interrupt
Dec 13 03:05:35 zfs01 kernel: nvme1: Missing interrupt
Dec 13 03:06:26 zfs01 kernel: nvme1: Missing interrupt
Dec 13 03:06:57 zfs01 kernel: nvme1: Missing interrupt
Dec 13 03:07:58 zfs01 kernel: nvme1: Missing interrupt
Jan 25 03:06:02 zfs01 kernel: nvme1: Missing interrupt
Jan 25 03:07:02 zfs01 kernel: nvme1: Missing interrupt
Feb 11 03:05:32 zfs01 kernel: nvme1: Missing interrupt
Feb 11 03:07:01 zfs01 kernel: nvme1: Missing interrupt
Feb 17 03:06:29 zfs01 kernel: nvme1: Missing interrupt

===

bzgrep "nvme1: Missing interrupt" /var/log/messages.0.bz2 ; grep "nvme1:
Missing interrupt" /var/log/messages
Jan 25 04:29:03 volume3 kernel: nvme1: Missing interrupt
Feb  4 04:04:45 volume3 kernel: nvme1: Missing interrupt
Feb 11 04:04:48 volume3 kernel: nvme1: Missing interrupt

Kind of interesting that zfs01 and volume03 have totally different customers,
usage patterns but have a collission of 2 dates :| 


> 2. Have you tried different instance types?

Yes. This issue is not manifesting itself on an r4.xlarge instance. Same ami on
r5.large: problems...

> 3. What sort of disk is this?

We use cloudformation/ansible to deploy these servers, so they are all kind of
identically configured (apart from sizing).  Both instances suffering from this
issue are 500GB+ EBS GP2 or IO disks with GPT and a zpool configured onto them. 

Please let me know if you would like to receive more information.

Kind regards,

Ruben

-- 
You are receiving this mail because:
You are the assignee for the bug.