#37 GC overhead limit exceeded

[Anonymous]

Hi JPM!

We have been using the asemon_logger for several years and have had very good experiences with it.
After upgrading our SAP ASE instances to ASE16 SP03 PL08 HF2, we also had to update the asemon_logger (current version 3.1.7).

We're monitoring 2 instances.

In the asemon_logger.sh the parameter -Xmx has already been increased several times up to 1024m. The monitoring of the larger instance runs smoothly. With the other one we keep getting this error:

Exception in thread "<instance>_BlockedP" java.lang.OutOfMemoryError: GC overhead limit exceeded
        at com.sybase.jdbc4.utils.BufferPool.makeBuffer(BufferPool.java:247)
        at com.sybase.jdbc4.utils.BufferPool.getBI(BufferPool.java:232)
        at com.sybase.jdbc4.timedio.InStreamMgr.setBuffer(InStreamMgr.java:498)
        at com.sybase.jdbc4.timedio.Dbio.doRead(Dbio.java:242)
        at com.sybase.jdbc4.timedio.InStreamMgr.readIfOwner(InStreamMgr.java:587)
        at com.sybase.jdbc4.timedio.InStreamMgr.doRead(InStreamMgr.java:319)
        at com.sybase.jdbc4.tds.TdsProtocolContext.getChunk(TdsProtocolContext.java:622)
        at com.sybase.jdbc4.tds.PduInputFormatter.readPacket(PduInputFormatter.java:239)
        at com.sybase.jdbc4.tds.PduInputFormatter.read(PduInputFormatter.java:72)
        at com.sybase.jdbc4.tds.TdsInputStream.read(TdsInputStream.java:88)
        at com.sybase.jdbc4.tds.TdsInputStream.readUnsignedByte(TdsInputStream.java:121)
        at com.sybase.jdbc4.tds.Tds.nextResult(Tds.java:3051)
        at com.sybase.jdbc4.tds.Tds.readCommandResults(Tds.java:4570)
        at com.sybase.jdbc4.tds.Tds.doCommand(Tds.java:4555)
        at com.sybase.jdbc4.tds.Tds.endTransaction(Tds.java:2695)
        at com.sybase.jdbc4.jdbc.SybConnection.commit(SybConnection.java:2063)
        at asemon_logger.CnxMgr$ArchCnxPool.putArchCnx(CnxMgr.java:126)
        at asemon_logger.Archive_ASE.saveValues_exec(Archive_ASE.java:510)
        at asemon_logger.Archive_ASE.saveValues(Archive_ASE.java:340)
        at asemon_logger.CollectorGeneric.saveValues(CollectorGeneric.java:475)
        at asemon_logger.CollectorGeneric.getMetrics(CollectorGeneric.java:284)
        at asemon_logger.Collector.run(Collector.java:380)

How can the memory required for the VM be estimated and why is 1024 m enough for one instance but not for the other one?

Incidentally, the BlockedP module is active in both cases.

Thank you very much in advance!

Best regards,
Kai.

[Anonymous]

Hi Kai,
from the java stack trace we see the problem occurs in jConnect (trying to acquire memory in InStreamMgr) Don't know if it is a Sybase bug

Do you have this problem very often ?
When the problem occurs you can execute manually the query of the BlockP collector and see if it retreives a lot of rows (asemon stores all rows in memory before inserting them in the asemon db)
The query :

select  
    time_blocked=0,
    blockedSpid=0,
    blockingSpid=0,

    L.SPID,
    L.KPID,
    L.DBID,
    ParentSPID,
    LockID,
    BlockedBy,
    BlockedState,
    Context,
    ObjectID,
    LockState,
    LockType,
    LockLevel,
    WaitTime,
    PageNumber,
    RowNumber,

    Usr=substring(suser_name(P.suid),1,30),
    Trn=substring(P.tran_name,1,30),
    Prg=substring(P.program_name,1,30),
    Prc=substring(     convert(varchar,P.id) ,1,30),
    Line=P.linenum,
    lckBase=substring(db_name(L.DBID),1,30),
    lckObjet=substring(object_name(L.ObjectID, L.DBID),1,30)


from master.dbo.monLocks L left outer join
     master.dbo.sysprocesses P on L.SPID=P.spid
where L.BlockedState is not null

Best regards
Jpm