On Thu, Dec 09, 2010 at 10:48:37AM -0800, Ying Han wrote: > On Thu, Dec 9, 2010 at 6:44 AM, Mel Gorman <mel@xxxxxxxxx> wrote: > > On Wed, Dec 08, 2010 at 05:23:24PM -0800, Andrew Morton wrote: > >> On Wed, 8 Dec 2010 16:36:21 -0800 Simon Kirby <sim@xxxxxxxxxx> wrote: > >> > >> > On Wed, Dec 08, 2010 at 04:16:59PM +0100, Johannes Weiner wrote: > >> > > >> > > Kswapd tries to rebalance zones persistently until their high > >> > > watermarks are restored. > >> > > > >> > > If the amount of unreclaimable pages in a zone makes this impossible > >> > > for reclaim, though, kswapd will end up in a busy loop without a > >> > > chance of reaching its goal. > >> > > > >> > > This behaviour was observed on a virtual machine with a tiny > >> > > Normal-zone that filled up with unreclaimable slab objects. > >> > > > >> > > This patch makes kswapd skip rebalancing on such 'hopeless' zones and > >> > > leaves them to direct reclaim. > >> > > >> > Hi! > >> > > >> > We are experiencing a similar issue, though with a 757 MB Normal zone, > >> > where kswapd tries to rebalance Normal after an order-3 allocation while > >> > page cache allocations (order-0) keep splitting it back up again. It can > >> > run the whole day like this (SSD storage) without sleeping. > >> > >> People at google have told me they've seen the same thing. A fork is > >> taking 15 minutes when someone else is doing a dd, because the fork > >> enters direct-reclaim trying for an order-one page. It successfully > >> frees some order-one pages but before it gets back to allocate one, dd > >> has gone and stolen them, or split them apart. > >> > > > > Is there a known test case for this or should I look at doing a > > streaming-IO test with a basic workload constantly forking in the > > background to measure the fork latency? > > We were seeing some system daemons(sshd) being OOM killed while > running in the same > memory container as dd test. I assume we can generate the test case > while running dd on > 10G of file in 1G container, at the same time running > unixbench(fork/exec loop)? > unixbench in a fork/exec loop won't tell us the latency of each individual operation. If order-1 is really a problem, we should see a large standard deviation between fork/exec attempts. A custom test of some sort is probably required. > > > >> This problem would have got worse when slub came along doing its stupid > >> unnecessary high-order allocations. > >> > >> Billions of years ago a direct-reclaimer had a one-deep cache in the > >> task_struct into which it freed the page to prevent it from getting > >> stolen. > >> > >> Later, we took that out because pages were being freed into the > >> per-cpu-pages magazine, which is effectively task-local anyway. But > >> per-cpu-pages are only for order-0 pages. See slub stupidity, above. > >> > >> I expect that this is happening so repeatably because the > >> direct-reclaimer is dong a sleep somewhere after freeing the pages it > >> needs - if it wasn't doing that then surely the window wouldn't be wide > >> enough for it to happen so often. But I didn't look. > >> > >> Suitable fixes might be > >> > >> a) don't go to sleep after the successful direct-reclaim. > >> > > > > I submitted a patch for this a long time ago but at the time we didn't > > have a test case that made a difference to it. Might be worth > > revisiting. I can't find the related patch any more but it was fairly > > trivial. > > If you have the patch, maybe we can give a try on our case. > I'll cobble one together early next week. -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom policy in Canada: sign http://dissolvethecrtc.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>