Changes between Version 1 and Version 2 of developers-meeting-20141014


Ignore:
Timestamp:
10/14/14 14:44:57 (6 years ago)
Author:
paul
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • developers-meeting-20141014

    v1 v2  
    1515== KIT == 
    1616 
    17 There was an issue with ATLAS SRM becoming overloaded due to activity from FTSv3 server.  It attempted to create/upload many files that already exist.  These operations failed because site-local configuration disables overwrite operations. 
     17There was an issue with ATLAS SRM becoming overloaded due to activity from FTSv3 server, reported as RT 8475.  The activity seems mostly with it attempting to create/upload many files that already exist.  These operations all failed because site-local configuration disables overwrite operations. 
    1818 
    19 100 concurrent transfers ... SRM got stuck and recovered badly from this 
     19The problem was that the SRM suffered from high CPU load.  It wasn't clear whether this load was from establishing the TCP/SSL connection or whether it was due to internal activity within dCache.  dCache log files contained messages indicating the database wasn't able to keep up with the load. 
    2020 
    21 SRM hit 100% CPU time; no transfers started.  Tried to overwrite existing files, which is not allowed  
     21During the meeting, Tigran sent some PostgreSQL configuration that should allow the database to provide improved performance. 
    2222 
    23 RT ticket 8475 
     23KIT's ATLAS instance is currently running dCache v2.6.33 
    2424 
    25 2.6.33 
    26  
    27 database optimisation based on Tigran's advise. 
    28  
    29 dCache team to talk to FTS guys. 
    30  
    31 Upload files that already exist? 
     25dCache team to talk to FTS guys to ask what is happening. 
    3226 
    3327Keep same connection -- if same user 
    3428 
    35 Now everyone seems to be using FTSv3; GridKa can switch off FTSv2 channels. 
     29Xavier said that it appears that everyone is now using FTSv3. 
    3630 
     31=== xroot issue === 
    3732 
    38 Noticed upgrade LHCb 2.6.34 to no confirmation that the problem is gone.  Lacking confirmation that Brazilian CA problem is fixed, too. 
     33KIT have upgraded their dCache to fix the xrootd problem.  They have received no feedback on whether this has fixed the problem. 
    3934 
     35They are also waiting for confirmation that the Brazilian CA problem is fixed. 
     36 
     37Paul said he would drop Philippe an email to try to prompt a response. 
    4038 
    4139= Support tickets for discussion =