Too many gaps in buffer synchronisation

[Former Member]

Hi all,

After referring to SAP Note 625242 and 575057, I'm checking the work process log to see which tables are affected.

This is from the log:

B ***LOG BS0=> too many gaps in buffer synchronisation [dbsync 2619 ]

B ***LOG BB1=> reset buffer NAM [dbntab 3282 ]

B ***LOG BB1=> reset buffer TABLP [dbbfe 4167 ]

B ***LOG BB1=> reset buffer TABL [dbtbxbuf1614 ]

B ***LOG BBD=> reset of buffer CUA , mode and status: InvalBeg [dbexpbuf2082 ]

B ***LOG BBD=> reset of buffer CUA , mode and status: InvalEnd [dbexpbuf2160 ]

B ***LOG FCK=> calendar buffer CALE was reset (mode: 3 ) [dbcalbuf4558 ]

=========================================================================

This is the content of the SAPNote:

The following message is found in the syslog:

BS0 memory overflow in the buffer synchronization

The trace file of the relevant work process contains the following messages:

B ***LOG BS0=> escalation level 2 for buffer synchronization reached

B Begin of notebook trace

B class = <n> table = <tabname>

B class = <n> table = <tabname>

B class = <n> table = <tabname>

=========================================================================

Comparing the log I obtained, and those shown in the SAP Note, it is different.

The problem now is I couldnt trace which buffered table is giving me all the repeating errors, and also, can anyone explain the error "Too many gaps in buffer synchronisation"?

Thanks.

[former_member184473 · Accepted solution]

Hello,

Are you using the minimum kernel mentioned in note [565057|http://service.sap.com/sap/support/notes/565057]?

Regards,

Eduardo Rezende

[Former Member · Accepted solution]

Hi,

May I know what did you mean by minimum kernel?

These are the info I have:

kernel release 46D

kernel make variant 46D_EXT

DBMS client library OCI_920_64__OCI_7_API

DBSL shared library version 46D.00

Thanks.

[former_member184473 · Accepted solution]

Hello,

This note is only valid for 610 onwards.

These symptoms usually occur if huge amounts of update operations are performed on buffered tables. Buffering should be undone on tables where write accesses dominate the read accesses.

If the problem re-occur immediately run report RSDBBUFF -> Statistics -> Synchronisation -> DDLOG records, change "Start time" to the respective time the problems started and increase "max. records" to a sufficient high number. Display the list and save it (unconverted).

The reason for the problem is described in more detail in note [14740|http://service.sap.com/sap/support/notes/14740].

Also, which are the current values for the parameters:

- rdisp/bufrefmode

- /rdisp/bufreftime

- zcsa/calendar_ids

- values from line Calendar on transaction ST02

Regards,

Eduardo

[Former Member · Accepted solution]

Hi Eduardo,

May I know what should I be looking at in the DDLOG entries?

I can see the same few tables (of TABLP) and DBDIFF (of TABL) dominating the list.

These are the current parameter settings.

ParameterName

rdisp/bufrefmode

Short description(Engl) Buffer Refresh Mode

Appl. area Dispatcher and Task Handler

ParameterTyp Special character strings

Changes allowed Change permitted

Valid for oper. system All operating systems

DynamicallySwitchable

Same on all servers

Special char. string sendon,sendoff,exeauto,exeoff

Separator ,

Dflt value sendon,exeauto

ProfileVal sendon,exeauto

Current value sendon,exeauto

ParameterName

rdisp/bufreftime

Short description(Engl) Refresh Interval in Distributed Systems

Appl. area Dispatcher and Task Handler

ParameterTyp Time value

Changes allowed Change permitted

Valid for oper. system All operating systems

Minimum 0

Maximum 999,999,999

DynamicallySwitchable

Same on all servers

Dflt value 120

ProfileVal 120

Current value 120

ParameterName

zcsa/calendar_ids

Short description(Engl) Directory Entries in the Calendar

Appl. area General system

ParameterTyp Integer value

Changes allowed Change permitted

Valid for oper. system All operating systems

Minimum 0

Maximum 0

DynamicallySwitchable

Same on all servers

Dflt value 200

ProfileVal 200

Current value 200

Calendar 100.00 488 478 100.00 200 200 100.00 0 139

Edited by: e.castro on Aug 20, 2011 2:26 AM

Edited by: e.castro on Aug 20, 2011 2:27 AM

[Former Member · Accepted solution]

Hi Eduardo,

Thanks. Sorry for the mess in previous message.

May I know what should I be looking at in the DDLOG entries?

I can see the same few tables (of TABLP) and DBDIFF (of TABL) dominationg the list.

Also, these are the parameter settings:

ParameterName

rdisp/bufrefmode

Short description(Engl) Buffer Refresh Mode

Appl. area Dispatcher and Task Handler

ParameterTyp Special character strings

Changes allowed Change permitted

Valid for oper. system All operating systems

DynamicallySwitchable

Same on all servers

Special char. string sendon,sendoff,exeauto,exeoff

Separator ,

Dflt value sendon,exeauto

ProfileVal sendon,exeauto

Current value sendon,exeauto

ParameterName

rdisp/bufreftime

Short description(Engl) Refresh Interval in Distributed Systems

Appl. area Dispatcher and Task Handler

ParameterTyp Time value

Changes allowed Change permitted

Valid for oper. system All operating systems

Minimum 0

Maximum 999,999,999

DynamicallySwitchable

Same on all servers

Dflt value 120

ProfileVal 120

Current value 120

ParameterName

zcsa/calendar_ids

Short description(Engl) Directory Entries in the Calendar

Appl. area General system

ParameterTyp Integer value

Changes allowed Change permitted

Valid for oper. system All operating systems

Minimum 0

Maximum 0

DynamicallySwitchable

Same on all servers

Dflt value 200

ProfileVal 200

Current value 200

Calendar 100.00 488 478 100.00 200 200 100.00 0 139

Edited by: e.castro on Aug 20, 2011 2:33 AM

[former_member184473 · Accepted solution]

Hello,

The 3 parameters (rdisp/bufrefmode, /rdisp/bufreftime and zcsa/calendar_ids) seems to be correct.

Also, from your description and affected objects are: NAM, TABLP, TABL, CUA and CALE.

Regards,

Eduardo

[Former Member · Accepted solution]

Hi Eduardo,

Would my next step of action is to change the tables that appeared in the DDLOG (at very high frequency) from buffered to non-buffered mode via SE11?

Thanks.

[former_member184473 · Accepted solution]

Hello,

You can also use report RSDBBUFF.

Regards,

Eduardo