[OpenAFS-devel] Fileserver fast start options - what effect?

[Hartmut Reuter]

Hello Harald,

these were my additions. They do the following:

--enable-fast-restart

lets the salvager exit immediately whenever it is called without 
arguments as happens after a chrash. The fileserver had to be modified 
to allow attachment of volumes in state V_inUse (actually these lines 
are in volume.c). I made this change because also at our site the 
salvager took very long to end up with only having updated the 
uniquifier in the volinfo-file which seem not to be worth such an delay.

Of course you can still salvage by "bos salvage server partition volume" 
or by "bos salvage server partition" or even by "bos salvage server 
-all". For NAMEI-fileservers salvaging single volumes is really fast 
because all data belonging to the volume group are uder a single 
directory. Therefore it makes sense to run a script to salvage each 
RW-volume one after the other instead of salvaging the partition because 
the fileserver has not to be stopped.

--enable-bitmap-later

could be enabled always. It reads the vnode-files to create the bitmap 
only when the 1st time a new vnode is going to be allocated. This makes 
the startup of the fileserver a little faster. I can't see any 
disadvantage of this option.

Both options are based on code I developed for our MR-AFS fileservers. 
Because we do not run any OpenAFS fileserver this code could not be 
tested extensively in OpenAFS by our site, but I heard theat other sites 
are running fileservers with these options.

Hartmut

Harald Barth schrieb:
> Lastest salvage took far too long (hours), and I stumbled over the
> options --enable-bitmap-later and --enable-fast-restart. As it was not
> very difficult to build the server by writing the options, I would
> appreciate more information from the author(s) what these actuallu do
> and don't do and if there is any more configuration required at
> startup. The amount of information found by for example google
> (search: enable-fast-restart) is astonishing low. Then I have some
> questions to folks actually using this (I hope there are): When _do_
> you run salvage? How do you detect volumes that are Off-line? What
> options do you use at fileserver startup? Of course, other
> observations, tricks concerning fast fileserver startup are welcome,
> too.
> 
> Thanks,
> Harald.
> 
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel


-- 
-----------------------------------------------------------------
Hartmut Reuter                           e-mail reuter@rzg.mpg.de
					   phone +49-89-3299-1328
RZG (Rechenzentrum Garching)               fax   +49-89-3299-1301
Computing Center of the Max-Planck-Gesellschaft (MPG) and the
Institut fuer Plasmaphysik (IPP)
-----------------------------------------------------------------