[platform-releng-dev] webmaster has addressed CVS timeout issues, rebuil

[Kim Moir <Kim_Moir@xxxxxxxxxx>]

Denis has adjusted NFS settings on the
nodes.  He thinks this may have been the source of our CVS checkout
problems earlier this week. 

Btw, I'm running a test build now for
the Equinox team.  If this test build is successful and the tests
run, I'll schedule an integration build for later this afternoon.

Kim


----- Forwarded by Kim
Moir/Ottawa/IBM on 03/13/2008 11:14 AM -----

bugzilla-daemon@xxxxxxxxxxx

03/13/2008 10:20 AM

To	Kim Moir/Ottawa/IBM@IBMCA
cc
Subject	[Bug 220526] I/O and 'file not found' errors from NFS server

https://bugs.eclipse.org/bugs/show_bug.cgi?id=220526
 
Product/Component: Community / Servers

Denis Roy <denis.roy@xxxxxxxxxxx> changed:

           What    |Removed    
                |Added
----------------------------------------------------------------------------
                 CC|    
                     
 |kmoir@xxxxxxxxxx
             Status|NEW      
                  |RESOLVED
         Resolution|          
                 |FIXED




--- Comment #3 from Denis Roy <denis.roy@xxxxxxxxxxx>  2008-03-13
10:20:10 -0500 ---
Ugh, I can be such a n00b sometimes.

A long, long time ago I set an option to return I/O Errors if an NFS timeout
occurred, instead of trying indefinitely, to avoid killing the nodes when
the
NFS server lagged. However, I misread the NFS docs and had the NFS timeout
set
*lower* than the default, thinking I had set it higher.  This means
I/O errors
were being returned at a much lower threshold than the default.

I have increased the timeout to something much, much higher, so these I/O
Errors should decrease dramatically, if not be eliminated altogether.

Kim, this could have been the reason (or a contributing factor to) your
CVS
checkouts were failing.  In my tests yesterday, I got one I/O Error
while
checking out the platform code.

None of this solves the kernel error from comment 1, but I don't think
it
actually has any impact on performance or data integrity, as the TCP packet
becomes broken,  the client will initiate a retry.  At any rate,
the symptom of
comment 1 happens about 4 times per day, so it's not as urgent to fix as
was
the I/O Error problem.

I'll close this as fixed.  Apologies for the boo boo.  Please
reopen if you get
more I/O errors and I'll tweak the timeout threshold some more.


-- 
Configure bugmail: https://bugs.eclipse.org/bugs/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.