- Aug 15, 2016
-
-
Giuseppe Lo Presti authored
only the digits for the CASTOR major/minor version constants.
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
- Aug 12, 2016
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
- Aug 11, 2016
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
Sebastien Ponce authored
When garbage collecting a DataPool, we let only one GC run at a time, as they are all looking for the same data.
-
Sebastien Ponce authored
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
Sebastien Ponce authored
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
- Aug 10, 2016
-
-
Giuseppe Lo Presti authored
Better handling of offline diskservers: don't try to contact them until they come back, and refresh the list every 30 seconds as opposed to 5 minutes
-
Giuseppe Lo Presti authored
-
Sebastien Ponce authored
We used to have a single connection, which means a single thread in the back answering to async replies. Now a pool is used. By default, we use 3 connections, but this can be changed via the CEPH/NbConnections entry of the castor.conf
-
Sebastien Ponce authored
Removed usage of castor.conf entry DiskManager/PoolUserId that was redondant with the DataPools entries
-
Sebastien Ponce authored
Added computation of checksums in the xrootd plugin for asynchronous calls. As amazing as it is, it was not there !
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
The logic in this thread has caused a number of DDoS when the transfermanagerd is stuck and reconnecting makes things even worse. Last occurrence with c2alice (ALARM ticket raised on Sat Aug 6th). A different solution may need to be implemented in the future.
-
Giuseppe Lo Presti authored
-
- Aug 08, 2016
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
This is suspected to have created a memory leak (but not a cursor leak) as it has been observed during the hardware decommissioning campaign, when many nodes were offline and the reporting was heavily exercised (in particular in c2alice but also in c2public). The problem stopped manifesting itself when the old nodes were completely taken out.
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
- Aug 05, 2016
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
The fix is to not log the 'DUMP' level any longer (but log all the rest), as we cannot intervene on the gridftp log's contents.
-
Giuseppe Lo Presti authored
for read and for write operations
-
Giuseppe Lo Presti authored
Fixed by replacing the literal * with the literal '*' string.
-
- Aug 01, 2016
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
- Jul 29, 2016
-
-
Giuseppe Lo Presti authored
-
Giuseppe Lo Presti authored
-
- Jul 25, 2016
-
-
Giuseppe Lo Presti authored
-
- Jul 20, 2016
-
-
Steven Murray authored
Add unimplemented MHVTL Volume Statistics SCSI page ## Description Page 0x17 is not implemented in MHVTL. As it inherits from DriveT10000D, it should be overridden in order to return empty map of metrics values. See merge request !8
-