Coda File System

Re: Released Coda 6.0.3 (and RPC2 1.20)

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Fri, 17 Oct 2003 18:43:50 -0400
Forgot to hit 'list-reply' the first time around,

On Sat, Oct 18, 2003 at 12:29:28AM +0200, Ivan Popov wrote:
> > === soon to be merged in CVS, this is NOT in 6.0.3! ===
> 
> >     Coda servers use a vnode lookup table that is causing some 'growing
> >     pains' for larger servers. The re-allocation of such a table would
> >     fail when we reached around 100k files per volume. These tables will
> >     be replaced by a hash table which with moderate success allowed us
> >     to store about a million files in a single volume. This does require
> >     some incompatible RVM changes to the vnode structure.
> >
> >     As we're going to be breaking the vnode structure anyways, we will
> >     take the opportunity to move the per file SHA1 checksum into RVM.
> >     This should significantly reduce server CPU usage when checksums
> >     are enabled.
> 
> Have you some ideas about the migration path for upgrades to that level?
> It looks like it will not be possible to use old rvm contents with the new
> server binaries, right?
> Will it be possible to dump-and-load rvm between the old and the new
> format? Alternatively dump-reinit-restore for all volumes?

We have been throwing around several ideas.

o transparent during server startup.
    + easy 'up' migration.
    - adds crud to the code that will be carried around forever.
    - if anything goes wrong halfway there is no way back, and possibly
      no way forward either.
    - lookup tables remain large.
    - any long term RVM fragmentation issues are not resolved.

o off-line rebuild with norton.
    + can deal with backup volumes.
    - lookup tables remain large.
    - any long term fragmentation issues are not resolved.
    - backward migration will have to be implemented as well.

o off-line rebuild with norton-reinit -dump/-restore.
    + user can restore the identical dumpfile with the old norton to
      migrate back.
    + server RVM will be as fresh as possible with minimal
      fragmentation.
    - have to purge backup volumes.

> It soon can become an issue for those who want the bleeding edge features,
> but have already got some production data on Coda.

I know, I'm dealing with the same issues, having about 12 'production'
servers. At the moment I don't even know how these changes affect simple
things like server-server resolution...

Jan
Received on 2003-10-17 18:44:45