OK, why does the glibc rpm need to execute /sbin/service?

Jakub Jelinek jakub at redhat.com
Wed Jun 8 05:54:30 UTC 2011


On Tue, Jun 07, 2011 at 06:39:47PM -0400, Tom Horsley wrote:
>   Updating   : glibc-2.14-2.x86_64                                        3/106 
> Non-fatal POSTIN scriptlet failure in rpm package glibc-2.14-2.x86_64
> /usr/sbin/glibc_post_upgrade: While trying to execute /sbin/service child exited with exit code 1
> warning: %post(glibc-2.14-2.x86_64) scriptlet failed, exit status 1
> 
> Should whatever it is doing be using systemctl instead?

No, systemd just should be fixed.  telinit u isn't synchronous in systemd,
but what's worse is that while systemd is reexecing service just fails
instead of waiting a tiny bit until the request can be satisfied.
So it isn't a problem just for glibc_post_upgrade which does
telinit u; service sshd condrestart, but any other app which will try
service/systemctl very soon after telinit u invoked from another app.

	Jakub


More information about the users mailing list