Project:Toolchain/Glibc 2.28 porting notes/glusterfs and sshd

From Gentoo Wiki
Jump to:navigation Jump to:search
Unrelated to glibc #681586
This is a single report; so far nobody else has observed it.

The usual sys-libs/glibc upgrade procedure is to update @world, then use app-admin/restart-services to restart daemons with changed libraries (most after a glibc update) and then a reboot (does not have to follow immediately). Since my distfiles are on glusterfs, an umount/mount is needed after the glibc update to reload libraries.

root #emerge -uDNva1 @world
root #umount /usr/portage/distfiles
root #mount /usr/portage/distfiles
root #restart-services

Prepare for the reboot, update kernel if necessary...

root #reboot

With glibc 2.28 this caused sshd crashes on several machines. Even the umount /usr/portage/distfiles command terminated my ssh connection. It's probably a good idea to have app-admin/monit installed to automatically restart sshd after a crash.

Because of that, this procedure seems to work better for glibc 2.28:

root #emerge -uDNva1 @world --exclude sys-libs/glibc

Prepare everything for the reboot, including kernel update if necessary

root #emerge -av1 sys-libs/glibc
root #reboot

It continues even after reboot

This has worked fine for some time, but a week after this happened again on multiple machines. Rebuilding/upgrading to glusterfs 6.0 didn't help either. It's not reproducible on all machines.

It can be triggered by:

root #mount /usr/portage/distfiles
root #emerge -ev1 sys-cluster/glusterfs
root #restart-services
root #umount /usr/portage/distfiles