URL:
<http://savannah.cern.ch/support/?119348>
Summary: Root reports an error while unzipping the buckets
fetched via xroot
Project: XROOTD
Submitted by: ljanyst
Submitted on: 2011-02-21 14:19
Category: None
Priority: 5 - Normal
Severity: 3 - Normal
Status: Need Info
Privacy: Public
Assigned to: ljanyst
Originator Email:
Open/Closed: Open
Discussion Lock: Any
Operating System: None
_______________________________________________________
Details:
Root reports unzipping errors when fetching data from an xrootd site. The
problem occurs when the connection is iddle for around 30 seconds or more.
Client version 20101014-1447 hangs for 5 minutes and then reports the
unzipping error. Newer client just hangs.
The client does the following operations:
1) file opening: it does some synchronous read requests to download the zip
header, the root file header and so on.
2) object reading: an async request to read 137216 bytes @719360
After a brief investigation it looks like the problem is that the answer to
the async request never comes when reading from glcc87.ucllnl.org and this
leeds to the following:
1) in the older version of ROOT (assuming that you haven't changed the
defaults):
a) the request timeout exceeds
b) the transaction timeout exceeds
c) the read operation fails and returns to ROOT
d) either root ignores the read error and proceeds to bucket unzipping or
the xrootd client doesn't signal the read error correctly, and as a result
ROOT tries to unzip some uninitialized memory hence the unzipping errors (a
bug that needs fixing, but not an urgent one)
2) in the newer versions:
a) the request timeout exceeds
b) there is still plenty of time to retry the read (the default
transaction timeout has been increased from 5 minutes to 8 hours) so the read
is retried and we're back at (a)
Now, this problem happens only when trying to read from glcc87.ucllnl.org, I
could not see any issue while reading from castor or an xroot service running
on my local PC, so I would assume that the server never sends any response to
the async request.
To be confirmed with the site admin whether the requests are not sent
correctly, not processed correctly on the server site or lost in the network.
_______________________________________________________
Reply to this item at:
<http://savannah.cern.ch/support/?119348>
_______________________________________________
Message sent via/by LCG Savannah
http://savannah.cern.ch/
|