removing a revision (Re: [PATCH] Best Practices)

[Michael Price <mp...@atl.lmco.com>]

Scott Lamb writes:
 > I once made the mistake of adding an OpenLDAP configuration file with my 
 > root DN's password embedded. (Hashed in SSHA, but still.) So tweaking 
 > later isn't always enough - you might want the incorrect stuff to not be 
 > in the history at all. Since there's no way to remove a revision (short 
 > of dumping/loading, and then probably manually re-adding any following 
 > revisions), best to avoid that.

I was about to ask if there was an easy way to remove revisions short of
dumping and loading. Last night I had two .pgn files (containing chess
move lists) in a directory. I wanted to commit the small one (3KB), not
the large one (73MB). Of course, I end up typing the filename of the large
one. So am I stuck dumping and loading now to get rid of the enormous
file?

On a side note, I quickly ran 'top' to see how much memory was going to be
used while adding this large file. I was pleased to see that subversion's
memory use never really increased while the httpd process grew to 65MB and
stayed that way. If I add a 2GB file (which I didn't think to try until
just now), what's going to happen to httpd? It'll be a few hours until I
get to try it so if anyone cares to try first, let me know.

Michael



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

Re: removing a revision (Re: [PATCH] Best Practices)

[Karl Fogel <kf...@newton.ch.collab.net>]

Michael Price <mp...@atl.lmco.com> writes:
> I tried it over ra_dav. Since only the httpd process was growing with the
> size of the file I figured it was just a ra_dav issue and didn't try the
> local case. The resident set size of the subversion process was only a few
> megabytes and didn't change during the entire operation.
> 
> Would you like me to fill out the issue or will you do it?

Would you mind filling it out?  You're the "primary source" :-)...

-K

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

Re: removing a revision (Re: [PATCH] Best Practices)

[Michael Price <mp...@atl.lmco.com>]

Karl Fogel writes:
 > Okay, thanks.  This is broken, of course; memory use should be roughly
 > constant, with data arriving streamily.  But I think it's a problem we
 > sort of knew about, although we don't have an issue filed for it.
 > 
 > Were you doing this over ra_local or ra_dav?  Or did you try both?
 > Curious to know if it behaved the same...

I tried it over ra_dav. Since only the httpd process was growing with the
size of the file I figured it was just a ra_dav issue and didn't try the
local case. The resident set size of the subversion process was only a few
megabytes and didn't change during the entire operation.

Would you like me to fill out the issue or will you do it?

Michael


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

Re: removing a revision (Re: [PATCH] Best Practices)

[Karl Fogel <kf...@newton.ch.collab.net>]

Okay, thanks.  This is broken, of course; memory use should be roughly
constant, with data arriving streamily.  But I think it's a problem we
sort of knew about, although we don't have an issue filed for it.

Were you doing this over ra_local or ra_dav?  Or did you try both?
Curious to know if it behaved the same...

-K


Michael Price <mp...@atl.lmco.com> writes:
> OK. Created a 2GB file containing random data. Created a new
> repository. Checked out the repository. Added the 2GB file. Did the
> commit. If you try this at home... have a lot of disk space because you'll
> need to store:
> 
>   1. the original 2GB file
>   2. the 2GB temporary file in .svn/tmp
>   3. the portion of the 2GB file that gets added to the repo
>   4. the log files for the portion that gets added to the repo
> 
> On my machine, each process has a hard limit of 448MB data segment size
> and 64MB stack size.
> 
> The first time, everything was going along fine until I ran out of disk
> space :)
> 
> The second time, after fixing the disk space issue, apache ran until it
> reached its process memory limit and then stopped. It had committed 448MB
> of the file when it stopped.
> 
> So, don't commit a file larger than any memory limits on the server or it
> won't work.


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

Re: removing a revision (Re: [PATCH] Best Practices)

[Michael Price <mp...@atl.lmco.com>]

Karl Fogel writes:
 > > On a side note, I quickly ran 'top' to see how much memory was going to be
 > > used while adding this large file. I was pleased to see that subversion's
 > > memory use never really increased while the httpd process grew to 65MB and
 > > stayed that way. If I add a 2GB file (which I didn't think to try until
 > > just now), what's going to happen to httpd? It'll be a few hours until I
 > > get to try it so if anyone cares to try first, let me know.
 > 
 > Let us know what happens!

OK. Created a 2GB file containing random data. Created a new
repository. Checked out the repository. Added the 2GB file. Did the
commit. If you try this at home... have a lot of disk space because you'll
need to store:

  1. the original 2GB file
  2. the 2GB temporary file in .svn/tmp
  3. the portion of the 2GB file that gets added to the repo
  4. the log files for the portion that gets added to the repo

On my machine, each process has a hard limit of 448MB data segment size
and 64MB stack size.

The first time, everything was going along fine until I ran out of disk
space :)

The second time, after fixing the disk space issue, apache ran until it
reached its process memory limit and then stopped. It had committed 448MB
of the file when it stopped.

So, don't commit a file larger than any memory limits on the server or it
won't work.

Michael



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

Re: removing a revision (Re: [PATCH] Best Practices)

[Karl Fogel <kf...@newton.ch.collab.net>]

Michael Price <mp...@atl.lmco.com> writes:
> I was about to ask if there was an easy way to remove revisions short of
> dumping and loading. Last night I had two .pgn files (containing chess
> move lists) in a directory. I wanted to commit the small one (3KB), not
> the large one (73MB). Of course, I end up typing the filename of the large
> one. So am I stuck dumping and loading now to get rid of the enormous
> file?

Yup.  Sorry :-(.

> On a side note, I quickly ran 'top' to see how much memory was going to be
> used while adding this large file. I was pleased to see that subversion's
> memory use never really increased while the httpd process grew to 65MB and
> stayed that way. If I add a 2GB file (which I didn't think to try until
> just now), what's going to happen to httpd? It'll be a few hours until I
> get to try it so if anyone cares to try first, let me know.

Let us know what happens!

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