Kernel 5.4.0-1025-aws causes massive load probably NFS-related
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
linux-gcp-5.4 (Ubuntu) |
New
|
Undecided
|
Unassigned |
Bug Description
On Ubuntu 18.04.5 LTS (GNU/Linux 5.4.0-1025-aws x86_64)
We had this:
# uptime
12:44:48 up 2:03, 3 users, load average: 66.94, 66.35, 65.28
After the upgrade from
linux-image-
to
linux-image-
on 2020-09-24
This here is after going back to 5.3.0-1035-aws:
# uptime
13:51:42 up 54 min, 1 user, load average: 1.38, 1.31, 1.2
The processes that were mainly causing the load were around 30 php-fpm processes that also connected to an external database via mysqlnd_ms, which saw a massive increase in connections and drop in bandwidth.
iotop reported nearly no reads/writes on the disks, and memory consumption was normal. The system felt responsive via ssh.
I realise this is difficult to debug; and I can't really use our (productive) systems to gather more data. But I hope somebody might know something about this (or at least somebody else having the same problems will find this).
summary: |
- Kernel 5.4.0-1025-aws causes massive load + Kernel 5.4.0-1025-aws causes massive load probably NFS-related |
Addendum: These use NFS too. So it might be related to that...