2 Replies Latest reply on Mar 4, 2013 10:25 PM by Vladislav Grigorov

    Tableau Server C++ Exception: Bad Allocation

    Vladislav Grigorov

      Hello Tableau gurus!


      First of all I apologize for raising the issue again, but I tried as hard as I could to troubleshoot the problem myself, following the guidelines in similar posts in the forum and still hard luck!


      The issue itself: I have a Tableau Server (v. 7.0.12) with per user licensing based on a private cloud infrastructure with the following specifications: 2x Intel Xeon X5675 @3,07GHz precessors (six-cores per proc.), 24GB RAM, 64-bit OS Windows Server 2008 R2. Same virtual machine is holding a SQL database of an ERP system with about 50 users, so workload is significant. There are several workbooks based off data extracts, containing crosstabs with some 60-70k rows and about 20 columns. These are filtered by the user to a specific time period and then exported to Excel, and when users do so, they often get error of the sort "C++ Exception: Bad Allocation" or "Out of memory". Sometimes just hitting the Refresh button renders the view correctly and/or allows export of the data, at other occasions a restart of the Tableau Server is the only viable solution.


      What is done so far:

      1) Reduced the vizqlserver.session.expiry.timeout from the standard 30 min to 5 min.

      2) Activated Windows Performance Monitoring to capture Tableau Server processes private bytes (the logs are visualized in the attached packaged workbook).

      3) Doubled the number of processes of Tableau Server (VizQL=4, Application=4, Data Server=4, Backgrounder=2).

      4) Increased the number of VizQL processes to 6, and increased the size of the windows swap file from 20 to 30 GB.

      5) Today I also generated the Tableau Server log files, that I find a bit hard to decode, but I am attaching them also.


      From the current state of affairs I see only three possible steps that I may take to alleviate the problem:

      6) Split the workbooks into very purpose-specific single-sheet, single-extract reports - not really nice to force the users down this path, but may turn to be the only viable option.

      7) Pay for more RAM (and/or other resources, but I think RAM is crucial in this case) on the VM where Tableau Server is based - costs money, so I really want to keep this as a final resort option.

      8) Keep restarting Tableau Server and wait for v.8 to be released to see if it will resolve the issue (new OpenGL rendering, etc.) - costs only time and inconvenience.


      To close that long and exhausting post: I really hope you can point me to something I am missing, or did wrong, or something I did not try, because it looks to me I have exhausted my list of clever ideas, and I neither have my finger on the issue, not have a clear plan of resolution yet!