68873 - Bugzilla install doesn't work over NFS; mention this in README

Status: VERIFIED FIXED

(Bugzilla :: Bugzilla-General, defect)

Description

[Thomas Stromberg]

We had a RAID fail on our Sun that powers our intranet site, so our webroot is
actually NFS mounted from a FreeBSD 4.1 machine. 

The Lock() function  "processmail" script requires flock() to work properly,
which may not work completely under NFS. After debugging, it hit me that we were
still going over NFS. The symptom is that checksetup.pl will hang on:

Now regenerating the shadow database for all bugs.

This is probably not really a bug, but more of a "user beware", and probably
something to mention in the README file. Especially for people installing to
machines which rely on network attached storage.

Comment 1

[Stephan Niemz]

I have checked in which places Bugzilla uses flock:

- globals.pl, if an SQL log is desired (data/sqllog exists and is
  writable),
- importxml.pl and move.pl, only used for moving bugs from one database to
  another,
- processmail, for logging to data/maillog and for protection against other
  instances of the same script

I agree that this should be mentioned in the README file (adjusting
summary).  Nominating for 2.12, since this shouldn't be too much effort.
But maybe someone else could do this since I'm not a native English
speaker. (Thomas?)

Comment 2

[Thomas Stromberg]

Attachment: An attempt at noting that Bugzilla & Checksetup need flock() in the README, and what goes wrong if you don't.

Comment 3

[Chris Yeh]

it should probably be noted at the top of the file somewhere near the top. the 
diff is fine as is. taking ownership from tara.

Comment 4

[Chris Yeh]

checked in your patch, and also added blurb up top under the pre-requisite 
section.  thanks for pointing this out.

Comment 5

[Matthew Tuck [:CodeMachine]]

Just I thought, should I file a new bug about detecting this in checksetup.pl?

Verified the doc change on my local install.

One small nitpick is that you guys seem to be using one space after full stops,
whereas the standard in the README file seems to be two.

Comment 6

[Dave Miller [:justdave]]

In search of accurate queries....  (sorry for the spam)

Comment 7

[Lilly]

Has there been a patch for this yet? since it's mentioned in the README, 
hopefully there's a way to patch it. It's a real problem now to submit bugs to 
the database at all. On the submit_bug.cgi, the submission hangs. checksetup.pl 
hangs. Everything is hanging.. 

It's constantly trying to submit against the shadow database, and I haven't 
found a way to disable that (I've been trying for the last two days!).

Thanks.

Comment 8

[Jacob Steenhagen]

I'm pretty sure that when oldmail is taken out of bugzilla (bug 71552) this will
go away (as the shadow directory is only used by the old e-mail diff method).

Comment 9

[Dave Miller [:justdave]]

Moving closed bugs to Bugzilla product