| === dgadomski_ is now known as dgadomski | ||
| === henrix_ is now known as henrix | ||
| caribou | Hello, is allocated hugetlb pages accounted for in the outpu of free ? | 10:01 |
|---|---|---|
| apw | caribou, i would expect it to be | 10:05 |
| caribou | apw: that's what I thought | 10:05 |
| caribou | apw: got two kernel panics caused by OOM; 118 Gb of memory is blocked by hugepages out of 125Gb | 10:06 |
| apw | caribou, hmmm, what kernel version is that ? | 11:24 |
| caribou | apw: 4.4.0-34-generic #53~14.04.1-Ubuntu | 11:43 |
| apw | caribou, i've not heard of anything like that, but it would be worth checking the reserves to see how much is being held | 12:07 |
| caribou | apw: I'm reading the code to understand what parameters triggers the OOM | 12:09 |
| arges | caribou: who allocated and locked the hugepages? | 12:15 |
| caribou | arges: they're used for QEMU & OpenVswitch | 12:16 |
| caribou | arges: they leave 8Gb out of 125Gb for the O/S & the rest goes to hugepages; my suspiscion is that they were close to the OOM threshold & something came in to push them over the edge | 12:17 |
| caribou | since the OOM threshold is calculated on the basis of 125Gb | 12:18 |
| arges | caribou: did you get a crash dump and use kmem | 12:19 |
| arges | did you/ were you able to | 12:19 |
| caribou | arges: yep, go two of them & waiting for another one from a stable system | 12:20 |
| caribou | arges: oh, & I found something funny : an 'lshw' process using 650Mb of RSS, stucked in doing do_huge_pmd_anonymous_page | 12:21 |
| arges | maybe there are a lot of pages : ) | 12:21 |
| === JanC is now known as Guest23252 | ||
| === JanC_ is now known as JanC | ||
| smoser | hey. i need some help. https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1618572 | 15:09 |
| ubot5 | Launchpad bug 1618572 in linux (Ubuntu) "cloud-init failing to configure sources and additional repository" [Undecided,Incomplete] | 15:09 |
| smoser | it appears to me that this is regression in overlayfs coming in xenial kernel | 15:10 |
| apw | smoser, what makes you think it is the overlayfs ? | 15:10 |
| _ami_ | wow, finally someone speaks! | 15:14 |
| smoser | apw, i do. | 15:18 |
| smoser | i could not reproduce it on a simple overlay mount | 15:19 |
| smoser | but i cuold reproduce it in 2 ways (apt-key add and a python program) | 15:19 |
| smoser | i started with a trusty system | 15:20 |
| smoser | and installed each of the linux-virtual-lts-utopic kernels, booted into them with overlay root and tried recreate | 15:21 |
| smoser | the failure starts in xenial kernel i'm pretty sure. | 15:21 |
| smoser | apw, if you need more info on how to recreate i can put it together. | 15:22 |
| tdaitx | apw, infinity: hey there! it seems the linux powerpc/ppc64el asm/ioctls.h is missing a include for asm-generic/ioctls.h (it's there for the other archs) | 15:43 |
| tdaitx | and I just noticed that the same happens for termbits.h | 15:44 |
| === neoark is now known as Guest3396 | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!