suggestions for large project (90K files, 930MB) - slow svn status

[Tim Harvey <ti...@yahoo.com>]

Greetings,

I recently setup a subversion server and have been very happy with it.  I'm
going to be working on a large project that has a tree of over 90K files making
up about 930MB.  While I'll only be working in one or two subdirs, those dirs
are at the top of the tree.  The project looks something like this:

projectA
  trunk
    |- linux-2.0.x
    |- linux-2.4.x
    |- linux-2.6.x
    |- user

I'll be working a bit in the linux-2.4.x tree, and a lot in the user tree, but
will need to be performing my builds from the trunk.  The linux-2.0.x and
linux-2.6.x trees are not really needed, but I would like to keep them in the
repository so that I can keep them up to date with new releases as I may use
them in the future.  My work will be modifications to 'projectA', however
'projectA' will get peridic releases that I will want to merge in from time to
time.

I ran a test where I created an FSFS repository and imported the tree.  Doing a
'svn status' at the top of the tree takes 4 minutes, which is going  to make it
fairly unusable for me.  Does this 4 minutes sound right?  Are there any
suggestions for how to layout a repository for this project?

I'm using a garden variety 1GHz x86 system with an IDE RAID5 array as my
server.  I'm currently using the svn schema.

Thanks for any suggestions

Tim

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: suggestions for large project (90K files, 930MB) - slow svn status

[Ben Collins-Sussman <su...@collab.net>]

On Dec 13, 2004, at 5:54 PM, Tim Harvey wrote:
>
> I ran a test where I created an FSFS repository and imported the tree. 
>  Doing a
> 'svn status' at the top of the tree takes 4 minutes, which is going  
> to make it
> fairly unusable for me.  Does this 4 minutes sound right?

Sure, why not?  How long would you expect 90,000 stat() calls to take?  
That's about 375 stat() calls per second.  That seems pretty good, no?

Remember that 'svn up' and 'svn commit' do the same thing.  So be 
prepared for the same pain if you run those commands from the top of 
such a huge working copy.

I certainly wouldn't expect CVS to behave any differently here...

>
> Thanks for any suggestions
>

My suggestion is that you not run 'svn status' from the top of such a 
huge working copy.  :-)

Work in smaller pieces... check out smaller sections of the repository. 
  Run 'svn status' and 'svn commit' on directories lower-down in the 
hierarchy when you can.  I imagine the *only* time you would run 'svn 
status' or 'svn commit' is when you absolutely need to commit a whole 
bunch of spread-out changes as a single changeset.


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: suggestions for large project (90K files, 930MB) - slow svn status

[Ni...@sybari.com]

> I'm using a garden variety 1GHz x86 system with an IDE RAID5 array as my
> server.  I'm currently using the svn schema.
Out of curiosity, what OS and file system?

Regards,
Nick G.



---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org