[16:33] <Odd_Bloke> cjwatson: wgrant: When we `mount -t (overlay|overlayfs) ...` in a build, the appropriate kernel module is loaded (despite modprobe not being permitted).  Will that module remain loaded for future builds on the host?
[16:37] <cjwatson> Odd_Bloke: On virtual builders, no; on non-virtual builders, yes.
[16:37] <cjwatson> (Well, until the host is rebooted.)
[16:40] <smoser> so virtual builders launch a new vm and build you inside a chroot
[16:40] <cjwatson> Correct.
[16:40] <smoser> and non-virtual builders just build you inside a chroot
[16:40] <cjwatson> Again correct.
[16:41] <smoser> ok, that makes sense.
[16:41] <cjwatson> This is another of the large class of reasons we want to get away from non-virtual builders.
[16:42] <smoser> yeah.
[16:42] <smoser> an lxc container i guess would plug the loading of module by mount kernel call
[16:42] <smoser> by denying the mount
[16:42] <smoser> but we need that in lots of cases
[16:43] <cjwatson> There's also very little benefit to using LXC for us given the now-preferred model of doing everything in a fresh VM anyway.
[16:58] <smoser> true story