Date: Tue, 10 May 2022 20:31:35 -0700 From: Mark Millard <marklmi@yahoo.com> To: Jan Mikkelsen <janm@transactionware.com>, Pete Wright <pete@nomadlogic.org> Cc: freebsd-current <freebsd-current@freebsd.org> Subject: Re: Chasing OOM Issues - good sysctl metrics to use? Message-ID: <A3D29E7A-62C1-492F-9631-06437C17B264@yahoo.com> In-Reply-To: <D429A8ED-011A-4E67-9726-C49937861CCD@yahoo.com> References: <83A713B9-A973-4C97-ACD6-830DF6A50B76.ref@yahoo.com> <83A713B9-A973-4C97-ACD6-830DF6A50B76@yahoo.com> <a5b2e248-3298-80e3-4bb6-742c8431f064@nomadlogic.org> <94B2E2FD-2371-4FEA-8E01-F37103F63CC0@yahoo.com> <0fcb5a4a-5517-e57b-2b69-4f3b3b10589a@nomadlogic.org> <DD98C932-A07F-4097-AE7F-D9CEF0BB6AEE@yahoo.com> <f43d7276-3718-df89-cbf0-5c1ef3d67e77@nomadlogic.org> <f00ccd1f-b6f6-bb00-f0a7-2f760c8953a0@nomadlogic.org> <464ED220-0DE4-4D2F-9DA2-AFD00D8D42B7@yahoo.com> <446d5913-a8c2-7dd0-860b-792fa9fe7c5b@nomadlogic.org> <33B740AA-A431-49CB-9F27-50B8C49734A2@yahoo.com> <3C5C183F-1471-4139-A53C-0B3815CFC25E@yahoo.com> <75C02C8C-6A5E-4E19-AC7D-B5DB704E8F16@transactionware.com> <C992DE63-AE7B-47F7-B679-B76D480AC0B1@yahoo.com> <D429A8ED-011A-4E67-9726-C49937861CCD@yahoo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 2022-May-10, at 17:49, Mark Millard <marklmi@yahoo.com> wrote:
> On 2022-May-10, at 11:49, Mark Millard <marklmi@yahoo.com> wrote:
>=20
>> On 2022-May-10, at 08:47, Jan Mikkelsen <janm@transactionware.com> =
wrote:
>>=20
>>> On 10 May 2022, at 10:01, Mark Millard <marklmi@yahoo.com> wrote:
>>>>=20
>>>> On 2022-Apr-29, at 13:57, Mark Millard <marklmi@yahoo.com> wrote:
>>>>=20
>>>>> On 2022-Apr-29, at 13:41, Pete Wright <pete@nomadlogic.org> wrote:
>>>>>>=20
>>>>>>> . . .
>>>>>>=20
>>>>>> d'oh - went out for lunch and workstation locked up. i *knew* i =
shouldn't have said anything lol.
>>>>>=20
>>>>> Any interesting console messages ( or dmesg -a or =
/var/log/messages )?
>>>>>=20
>>>>=20
>>>> I've been doing some testing of a patch by tijl at FreeBSD.org
>>>> and have reproduced both hang-ups (ZFS/ARC context) and kills
>>>> (UFS/noARC and ZFS/ARC) for "was killed: failed to reclaim
>>>> memory", both with and without the patch. This is with only a
>>>> tiny fraction of the swap partition(s) enabled being put to
>>>> use. So far, the testing was deliberately with
>>>> vm.pageout_oom_seq=3D12 (the default value). My testing has been
>>>> with main [so: 14].
>>>>=20
>>>> But I also learned how to avoid the hang-ups that I got --but
>>>> it costs making kills more likely/quicker, other things being
>>>> equal.
>>>>=20
>>>> I discovered that the hang-ups that I got were from all the
>>>> processes that I interact with the system via ending up with
>>>> the process's kernel threads swapped out and were not being
>>>> swapped in. (including sshd, so no new ssh connections). In
>>>> some contexts I only had escaping into the kernel debugger
>>>> available, not even ^T would work. Other times ^T did work.
>>>>=20
>>>> So, when I'm willing to risk kills in order to maintain
>>>> the ability to interact normally, I now use in
>>>> /etc/sysctl.conf :
>>>>=20
>>>> vm.swap_enabled=3D0
>>>=20
>>> I have been looking at an OOM related issue. Ignoring the actual =
leak, the problem leads to a process being killed because the system was =
out of memory. This is fine. After that, however, the system console was =
black with a single block cursor and the console keyboard was =
unresponsive. Caps lock and num lock didn=E2=80=99t toggle their lights =
when pressed.
>>>=20
>>> Using an ssh session, the system looked fine. USB events for the =
keyboard being disconnected and reconnected appeared but the keyboard =
stayed unresponsive.
>>>=20
>>> Setting vm.swap_enabled=3D0, as you did above, resolved this =
problem. After the process was killed a perfectly normal console =
returned.
>>>=20
>>> The interesting thing is that this test system is configured with no =
swap space.
>>>=20
>>> This is on 13.1-RC5.
>>>=20
>>>> This disables swapping out of process kernel stacks. It
>>>> is just with that option removedfor gaining free RAM, there
>>>> fewer options tried before a kill is initiated. It is not a
>>>> loader-time tunable but is writable, thus the
>>>> /etc/sysctl.conf placement.
>>>=20
>>> Is that really what it does? =46rom a quick look at the code in =
vm/vm_swapout.c, it seems little more complex.
>>=20
>> I was going by its description:
>>=20
>> # sysctl -d vm.swap_enabled
>> vm.swap_enabled: Enable entire process swapout
>>=20
>> Based on the below, it appears that the description
>> presumes vm.swap_idle_enabled=3D=3D0 (the default). In
>> my context vm.swap_idle_enabled=3D=3D0 . Looks like I
>> should also list:
>>=20
>> vm.swap_idle_enabled=3D0
>>=20
>> in my /etc/sysctl.conf with a reminder comment that the
>> pair of =3D0's are required for avoiding the observed
>> hang-ups.
>>=20
>>=20
>> The analysis goes like . . .
>>=20
>> I see in the code that vm.swap_enabled !=3D0 causes
>> VM_SWAP_NORMAL :
>>=20
>> void
>> vm_swapout_run(void)
>> {
>>=20
>> if (vm_swap_enabled)
>> vm_req_vmdaemon(VM_SWAP_NORMAL);
>> }
>>=20
>> and that in turn leads to vm_daemon to:
>>=20
>> if (swapout_flags !=3D 0) {
>> /*
>> * Drain the per-CPU page queue batches as a =
deadlock
>> * avoidance measure.
>> */
>> if ((swapout_flags & VM_SWAP_NORMAL) !=3D 0)
>> vm_page_pqbatch_drain();
>> swapout_procs(swapout_flags);
>> }
>>=20
>> Note: vm.swap_idle_enabled=3D=3D0 && vm.swap_enabled=3D=3D0 ends
>> up with swapout_flags=3D=3D0. vm.swap_idle. . . defaults seem
>> to be (in my context):
>>=20
>> # sysctl -a | grep swap_idle
>> vm.swap_idle_threshold2: 10
>> vm.swap_idle_threshold1: 2
>> vm.swap_idle_enabled: 0
>>=20
>> For reference:
>>=20
>> /*
>> * Idle process swapout -- run once per second when pagedaemons are
>> * reclaiming pages.
>> */
>> void
>> vm_swapout_run_idle(void)
>> {
>> static long lsec;
>>=20
>> if (!vm_swap_idle_enabled || time_second =3D=3D lsec)
>> return;
>> vm_req_vmdaemon(VM_SWAP_IDLE);
>> lsec =3D time_second;
>> }
>>=20
>> [So vm.swap_idle_enabled=3D=3D0 avoids VM_SWAP_IDLE status.]
>>=20
>> static void
>> vm_req_vmdaemon(int req)
>> {
>> static int lastrun =3D 0;
>>=20
>> mtx_lock(&vm_daemon_mtx);
>> vm_pageout_req_swapout |=3D req;
>> if ((ticks > (lastrun + hz)) || (ticks < lastrun)) {
>> wakeup(&vm_daemon_needed);
>> lastrun =3D ticks;
>> }
>> mtx_unlock(&vm_daemon_mtx);
>> }
>>=20
>> [So VM_SWAP_IDLE and VM_SWAP_NORMAL are independent bits
>> in vm_pageout_req_swapout.]
>>=20
>> vm_deamon does:
>>=20
>> mtx_lock(&vm_daemon_mtx);
>> msleep(&vm_daemon_needed, &vm_daemon_mtx, PPAUSE, =
"psleep",
>> vm_daemon_timeout);
>> swapout_flags =3D vm_pageout_req_swapout;
>> vm_pageout_req_swapout =3D 0;
>> mtx_unlock(&vm_daemon_mtx);
>>=20
>> So vm_pageout_req_swapout is regenerated after thata
>> each time.
>>=20
>> I'll not show the code for vm.swap_idle_enabled!=3D0 .
>>=20
>=20
> Well, with continued experiments I got an example of
> a hangup for which looking via the db> prompt did not
> show any swapping out of process kernel stacks
> ( vm.swap_enabled=3D0 was the context, so expected ).
> The environment was ZFS (so with ARC).
>=20
> But this was testing with vm.pageout_oom_seq=3D120 instead
> of the default vm.pageout_oom_seq=3D12 . It may be that
> let sit long enough things would have unhung (external
> perspective).
>=20
> It is part of what I'm experimenting with so we will see.
>=20
Looks like I might have overreacted, in that for my
current tests there can be brief periods of delayed
response, but things respond in a little bit.
Definately not like the hang-ups I was getting with
vm.swap_enabled=3D1 .
=3D=3D=3D
Mark Millard
marklmi at yahoo.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?A3D29E7A-62C1-492F-9631-06437C17B264>