| === himcesjf_ is now known as him-cesjf | ||
| oursland | Hi all, I get a build error when following the wiki on rebuilding the kernel. The error seems to be due to SPL/ZFS. The error and my build scripts are available: https://gitlab.com/snippets/1696811 Any advice on getting the kernel to build? | 05:41 |
|---|---|---|
| === ulysses_ is now known as ulysses | ||
| === zyga_ is now known as zyga | ||
| apw | oursland, are you building from a mainline builds source ? | 09:01 |
| apw | oursland, if so those don't even have zfs applied so you can only build those with do_zfs=false | 09:02 |
| apw | oursland, and tools always build into a 'one pre architecture' package, but the primary kernle build should build a linux-tools-*-<flavour> with links in it | 09:02 |
| frickler | we are having multiple servers going oom in a couple of days reproducibly after upgrading to 4.13-hwe kernels. they are running fine with 4.10-hwe (did so for a year and do again after booting with the older kernel again). has anyone seen similar things? this also happens with older 4.13-0.19, so likely not related to recent pti/meltdown stuff | 09:11 |
| apw | frickler, i do not think i have heard of anything like yet as yet ... coul dyou file a bug against the kernle please | 09:14 |
| apw | frickler, and put as much detail about how you are triggering that | 09:14 |
| frickler | apw: https://bugs.launchpad.net/ubuntu/+source/linux-hwe/+bug/1748408 I'm testing with upstream kernel now | 09:48 |
| ubot5 | Ubuntu bug 1748408 in linux-hwe (Ubuntu) "Servers going OOM" [Undecided,New] | 09:48 |
| apw | frickler, great, if you can find somewhere not affected that is always good | 09:56 |
| frickler | apw: well, 4.10 is fine, but I don't know whether a bisect from 4.10 to 4.13 would be feasible | 10:00 |
| apw | frickler, well if you are able to test the mainline builds if they are full featured enough | 10:01 |
| apw | you might well be able to work through 4.11 4.12 and 4.13 mainline | 10:01 |
| apw | to see which of those work, then we move to the -rc builds, and then we can bisect in that gap | 10:01 |
| frickler | apw: at least one system is running now with 4.15, will need to wait a couple of hours to see how it behaves | 10:02 |
| apw | frickler, i've passed the bug on to someone who can help with a bisect when you get to that point | 10:13 |
| frickler | apw: I just noticed that the mainline kernels are lacking aufs support, so docker won't run properly, giving less confidence in my testing results for these kernels. is it missing intentionally? | 13:31 |
| apw | frickler, they are mainline kernels, they can only have things which are mainline | 14:03 |
| frickler | apw: yeah, I found out in the meantime that aufs has been removed after having long been deprecated, working to replace that now | 14:14 |
| apw | aufs was never upstream, so it isn't in any of the mainline builds | 14:14 |
| frickler | apw: hmm, strange, not sure why it is being used by default then in our setup. but anyway, can't be too bad to move away from it. would be even nicer if it turned out to have caused my original issue ;) | 14:18 |
| apw | heh | 14:18 |
| oursland | apw: do_zfs=false fixed my problem. Thanks! | 21:21 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!