main difference between the system where oom happens is: - Single Xeon => no OOM - Dual Xeon / NUMA => OOM
both 64GB mem. Am 07.10.2016 um 11:33 schrieb Holger Hoffstätte: > On 10/07/16 09:17, Wang Xiaoguang wrote: >> Hi, >> >> On 10/07/2016 03:03 PM, Stefan Priebe - Profihost AG wrote: >>> Dear Wang, >>> >>> can't use v4.8.0 as i always get OOMs and total machine crashes. >>> >>> Complete traces with your patch and some more btrfs patches applied (in >>> the hope in fixes the OOM but it did not): >>> http://pastebin.com/raw/6vmRSDm1 >> I didn't see any such OOMs... >> Can you try holger's tree with my patches. > > They don't really apply to either 4.4.x (because it has diverged too > much by now) or 4.8.x because of the initial dedupe support which came > in as part of 4.9rc1 - there are way too many conflicts all over the > place and merging them manually took way too much time. > It would be useful if you could rebase your patches to for-next. > > Stefan, have you tried setting THP to 'madvise' or 'never'? > Try 'echo madvise > /sys/kernel/mm/transparent_hugepage/enabled' > or boot with transparent_hugepage=madvise (or never) kernel flag. > I have no idea if it will help, but it's worth a try. > > -h > -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html