Comment # 26 on bug 1107617 from LTC BugProxy

------- Comment From nevdull@us.ibm.com 2019-05-06 18:19 EDT-------
SUSE - I have access to our 64TB , 1536 cpu machine this week.  It is currently
running

Linux linux-bpkd 4.12.14-193-default #1 SMP Tue Apr 9 13:29:32 UTC 2019
(85c4156) ppc64le ppc64le ppc64le GNU/Linux

If you have a modified udev you'd like me to try, or additional data you'd like
me to collect, let me know.

With the release-provided udev, the default timeout of 90 sec is currently not
enough to discover the console, (LVM) root, or (LVM) swap before dropping into
emergency mode, but changing the timeout to 5 min (600 seconds) seems to be
sufficient now.  Previous releases, however, had required a timeout of 1800 or
2400 seconds because we seemed to be encountering a thundering herd (see SUSE
1103094) that resulted in unnaturally long delays.  It's not clear why the
behavior changed.  While those excesses are no longer the default, I discovered
they can be artificially induced for test purposes in the current release by
specifying udev.children-max=200000 as a boot argument.

What is the default timeout in the new systemd?  While a timeout of 600 seconds
or more may be overkill for laptops and even many rack machines, we may wish to
have this behavior documented for installation of large configurations (if it
isn't already.)