Postgres Pro
Downloads
Home   >   mailing lists

Thread: High kswapd

High kswapd

From
Don Seiler
Date:
Good morning,

Ubuntu 16.04.6 LTS
PostgreSQL 9.6.5

On one of our database servers, we're regularly seeing kswapd at the top of "top" output, regularly using over 50 %CPU. We should have well over 80GB of available memory according to "free -m".

# free -m
              total        used        free      shared  buff/cache   available
Mem:         125910       41654         820         857       83435       82231
Swap:           511         448          63

We've already got vm.swappiness and vm.zone_reclaim_mode set to 0, and NUMA is disabled from what I can see:

# dmesg | grep -i numa
[    0.000000] No NUMA configuration found

We are using HugePages and things look good there as well. Curious what would be causing kswapd to run hot like it is, or if it is a red herring as I look into high CPU usage on this box (although, again, it is the single-highest CPU user).

--
Don Seiler
www.seiler.us

Re: High kswapd

From
Justin Pryzby
Date:
On Mon, Apr 13, 2020 at 09:34:23AM -0500, Don Seiler wrote:
> Good morning,
> 
> Ubuntu 16.04.6 LTS
> PostgreSQL 9.6.5
> 
> On one of our database servers, we're regularly seeing kswapd at the top of
> "top" output, regularly using over 50 %CPU. We should have well over 80GB
> of available memory according to "free -m".

Do you have THP enabled?  Or KSM ?

tail /sys/kernel/mm/ksm/run /sys/kernel/mm/transparent_hugepage/khugepaged/defrag
/sys/kernel/mm/transparent_hugepage/enabled/sys/kernel/mm/transparent_hugepage/defrag
 

https://www.postgresql.org/message-id/20170718180152.GE17566%40telsasoft.com

-- 
Justin

Re: High kswapd

From
Don Seiler
Date:


On Mon, Apr 13, 2020 at 9:42 AM Justin Pryzby <pryzby@telsasoft.com> wrote:

tail /sys/kernel/mm/ksm/run /sys/kernel/mm/transparent_hugepage/khugepaged/defrag /sys/kernel/mm/transparent_hugepage/enabled /sys/kernel/mm/transparent_hugepage/defrag

# tail /sys/kernel/mm/ksm/run /sys/kernel/mm/transparent_hugepage/khugepaged/defrag /sys/kernel/mm/transparent_hugepage/enabled /sys/kernel/mm/transparent_hugepage/defrag
==> /sys/kernel/mm/ksm/run <==
0

==> /sys/kernel/mm/transparent_hugepage/khugepaged/defrag <==
1

==> /sys/kernel/mm/transparent_hugepage/enabled <==
always madvise [never]

==> /sys/kernel/mm/transparent_hugepage/defrag <==
always madvise [never]


--
Don Seiler
www.seiler.us

Re: High kswapd

From
Justin Pryzby
Date:
On Mon, Apr 13, 2020 at 09:46:22AM -0500, Don Seiler wrote:
> ==> /sys/kernel/mm/ksm/run <==
> 0

Was it off to begin with ?
If not, you can set it to "2" to "unshare" pages.

> ==> /sys/kernel/mm/transparent_hugepage/khugepaged/defrag <==
> 1

So I'd suggest trying with this disabled.

I don't know if I ever fully understood the problem, but it sounds like at
least in your case it's related to large shared_buffers, and hugepages, which
cannot be swapped out.

-- 
Justin

Re: High kswapd

From
Don Seiler
Date:
On Mon, Apr 13, 2020 at 9:58 AM Justin Pryzby <pryzby@telsasoft.com> wrote:
On Mon, Apr 13, 2020 at 09:46:22AM -0500, Don Seiler wrote:
> ==> /sys/kernel/mm/ksm/run <==
> 0

Was it off to begin with ?
If not, you can set it to "2" to "unshare" pages.

Yes we haven't changed this. It was already set to 0.
 

> ==> /sys/kernel/mm/transparent_hugepage/khugepaged/defrag <==
> 1

So I'd suggest trying with this disabled.

My understanding was that THP is disabled anyway. What would this defrag feature be doing now?
 
I don't know if I ever fully understood the problem, but it sounds like at
least in your case it's related to large shared_buffers, and hugepages, which
cannot be swapped out.

Basically the problem is our DB host getting slammed with connections (even with pgbouncer in place). We see the CPU load spiking, and when we check "top" we regularly see "kswapd" at the top of the list. For example, just now kswapd is at 72 %CPU in top. The next highest is a postgres process at 6.6 %CPU.

Our shared_buffers is set to 32GB, and HugePages is set to 36GB:

# grep Huge /proc/meminfo
AnonHugePages:         0 kB
HugePages_Total:   18000
HugePages_Free:     1897
HugePages_Rsvd:       41
HugePages_Surp:        0
Hugepagesize:       2048 kB 

Also FWIW this host is actually a VSphere VM. We're looking into any underlying events during these spikes as well.

Don.
--
Don Seiler
www.seiler.us