On Thu, 2014-06-12 at 20:00 -0700, Linda Walsh wrote:
lynn wrote:
On Thu, 2014-06-12 at 15:00 -0700, Linda Walsh wrote:
Are the autofs files the same between the two clients (suse+ubuntu)?
---- The clients are the same you said? Have you looked at how they are built (i.e. suse src rpm and similar for ubuntu -- see if they used same options)
does autofs use a cache file anywhere?
Mmm. I wonder. Not autofs as such, but both sssd and kerberos have their own cache. One problem in 13.1 is that systemd puts the root ticket under /var/run/user/0, whereas Kerberos needs it in /tmp. Of course, if root has not logged in, that directory will not exist. root will never have logged in on a client machine anyway so this is one of our systemd annoyances. The workaround is the krb5.conf file we quoted. The other cache are the ldbs maintained by sssd under /var/log/sss/db. The machine ticket cache for the automounter is also stored there but as sssd does it's own kerberos implementation, it's not necessary for that ticket to be in /tmp. Is it? Those work like nscd for getent. Forget to clear them, change something and nothing happens. Something else that narrows this down is that with the maps and nss as files (such as /etc/auto.master) the shares get mounted even if the file server was down at boot. This is why the suggestion to bugzilla against autofs seems wrong. But that doesn't get us away from an Ubuntu ws working out of the box against the same DC and fs. Sorry folks for rambling on. Sometimes thinking out loud will bring out the obvious I normally always always overlook. The m$ mob are pointing and clicking on the test domain ATM so everyone else is not welcome.
how about the autofs versions? Another thing to look at might be build options -- i.e. if the config files are the same, then something in the suse config is caching the fact that the host you want is missing in the suse version, but not in the ubuntu version.
Looking at the source and build options to see which ones might affect caching might be another step (all other things being the same, of which the systemd-boot graph says nothing). Is ubuntu running systemd?
^^^^^^^^^^^^^^^^^^^^^^^^^ systemd? ubuntu?
It's possible queries from autofs are hitting some cache on the suse system that doesn't exist on the ubuntu system ??
Just some ideas... now that the pretty pictures are out of the way...;-)
---- I think I could get a similar graph to the one you posted, I agree, it was pretty, but I'd have to do a bit more work.
As it is, I only could easily find the kernel time (attached)-- which ends up being about 1/3 as long using sysV and direct boot vs. systemd. I also read that for speed, some kernel devs suggest chucking udev -- apparently it is pretty slow compared to some alternatives (uboot+mdev).
I seem to remember ubuntu using some of those alternate boot methods maybe between caching differences and maybe a large variation in boot order and speed, ubuntu isn't hitting the same problem areas.
To compare the two -- you might turn autofs "off", then, manually start it on both systems and see if you see the same behaviors.
At least then we could rule out boot speed and maybe order, as well.
FWIW, I am still at samba 3.x, so most of your config is lost on me...(sigh)... sorry. The file server is exactly the same. The only (big) difference is active directory on the DC. Oh, and you don't have to put all that create mode 777 or whatever smb.conf nonsense any more.
Thanks for your input, L x -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org