http://bugzilla.opensuse.org/show_bug.cgi?id=1207807 http://bugzilla.opensuse.org/show_bug.cgi?id=1207807#c1 Alexander Krupp changed: What |Removed |Added ---------------------------------------------------------------------------- Found By|--- |Other See Also| |https://bugzilla.redhat.com | |/show_bug.cgi?id=879801 --- Comment #1 from Alexander Krupp --- There has been a similar report here https://bugzilla.redhat.com/show_bug.cgi?id=879801 , however it was closed WONTFIX. A workaround is proposed there: https://bugzilla.redhat.com/show_bug.cgi?id=879801#c13 The workaround does not solve the matter completely. -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1207807 http://bugzilla.opensuse.org/show_bug.cgi?id=1207807#c2 Alexander Krupp changed: What |Removed |Added ---------------------------------------------------------------------------- See Also| |https://bugzilla.redhat.com | |/show_bug.cgi?id=1694305 --- Comment #2 from Alexander Krupp --- https://bugzilla.redhat.com/show_bug.cgi?id=1694305 contains a hint about vmware page trimming -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1207807 http://bugzilla.opensuse.org/show_bug.cgi?id=1207807#c5 --- Comment #5 from Alexander Krupp --- Thank you for pointing out the issue including the trace-cmd. I am running at present with linux memory compaction enabled, but with VMware page trimming disabled as suggested in https://bugzilla.redhat.com/show_bug.cgi?id=1694305#c12 The issue did occur less frequently over the past hours, but it still happens. I can trigger it at will by issuing: echo 1 >/proc/sys/vm/compact_memory Result: the vmware guest freezes instantly for approx. 25s. The trace I uploaded was started ~ 2s before memory compaction was requested and it was stopped once the guest unfroze. I hope that helps somewhat. My kernel is Linux en-pc82 5.14.21-150400.24.41-default #1 SMP PREEMPT_DYNAMIC Fri Jan 13 08:55:22 UTC 2023 (1d4442d) x86_64 x86_64 x86_64 GNU/Linux however the problem has haunted me for a couple of months, and across multiple kernel versions already. -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1207807 http://bugzilla.opensuse.org/show_bug.cgi?id=1207807#c9 --- Comment #9 from Alexander Krupp --- cf. comment 6 -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1207807 http://bugzilla.opensuse.org/show_bug.cgi?id=1207807#c14 --- Comment #14 from Alexander Krupp --- Thank you for the suggestion Michal. I guess that running compaction in the background makes sense from a memory management view-point. I am running with vm.compaction_proactiveness=1 (instead of the standard value 20) for some time now and the issue has disappeared. I have also re-enabled the vmware page trimming to no ill effect. Even echo 1 >/proc/sys/vm/compact_memory does return now within about a second, most of the time, but it has not again taken ~ 30s, as before. I have created /etc/sysctl.d/vmware.conf with this setting. However, I do not know whether VMWare is the only (virtualization) software affected by this. If this affects more software and users, it may make sense to either change the default setting, or include this setting in some performance profile daemon settings for, e.g., "virtual-host"? Personally, I would be content with this bug being closed. -- You are receiving this mail because: You are on the CC list for the bug.

http://bugzilla.opensuse.org/show_bug.cgi?id=1207807 http://bugzilla.opensuse.org/show_bug.cgi?id=1207807#c16 --- Comment #16 from Vlastimil Babka --- (In reply to Alexander Krupp from comment #14)

However, I do not know whether VMWare is the only (virtualization) software affected by this. If this affects more software and users, it may make sense

I haven't heard of others than vmware. Certainly if it was affecting kvm or xen the same way, we would have heard and debugged that in upstream. (In reply to Alexander Krupp from comment #15)

RH disables compaction by default

https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/ html/8.4_release_notes/kernel_parameters_changes

Found this while reading up on what this parameter does.

I vote for the "principle of least surprise". As did RH.

But the important details is that this is for RHEL 8.4 with a kernel based AFAICS on 4.18, where upstream proactive compaction was merged later than that. Which suggests they backported the feature between 8.3 and 8.4 and in a minor service pack not updating the whole kernel, the principle of least surprise may indeed be important. If the feature comes with a new upstream kernel, usually the defaults are used unless there's a known problem. -- You are receiving this mail because: You are on the CC list for the bug.