|
Problem IC75870
|
Status: Closed |
GLOBAL SNAPSHOTS MAY CONSUME A LARGE AMOUNT OF FCM BUFFERS |
| product: |
DB2 FOR LUW / DB2FORLUW / 950 - DB2 |
| Problem description: |
In a multipartition environment, a large amount of FCM buffers
can be consumed by large global snapshots. One example of such
large global snapshots is, "db2 get
snapshot for locks on <dbname> global" when there are many
locks currently being held in the database.
db2pd -fcm hwm output would show a single application
consuming a lot of buffers |
| Problem Summary: |
****************************************************************
* USERS AFFECTED: *
* DPF environments *
****************************************************************
* PROBLEM DESCRIPTION: *
* Global snapshots collect monitoring information from all *
* database partitions and aggregate the result at the *
* coordinating partition. FCM buffers are required to send and *
* receive the monitoring information. The coordinating *
* partition can consume a large number of FCM buffers *
* equivalent to the size of the entire snapshot result. If *
* the database manager runs out of FCM buffers then node *
* failure is invoked to remedy the situation. This could lead *
* to applications receiving -1229 errors or losing connection *
* to the database. *
* *
* Note that monitoring tools like db2top issue global *
* snapshots. *
****************************************************************
* RECOMMENDATION: *
* Upgrade to DB2 v9.5 FP8 or later and enable flow control. *
**************************************************************** |
| Local Fix: |
The work around of setting DB2_MAX_GLOBAL_SNAPSHOT_SIZE should
help to prevent a single snapshot from consuming all FCM
buffers. Multiple concurrent large snapshots would still be a
problem though. |
| available fix packs: |
DB2 Version 9.5 Fix Pack 8 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 9 for Linux, UNIX, and Windows
DB2 Version 9.5 Fix Pack 10 for Linux, UNIX, and Windows
|
| Solution |
As of DB2 v9.5 fp8, the fix for this APAR can be enabled by
using the following registry variable setting:
db2set DB2_SYSTEM_MONITOR_SETTINGS=ENABLE_FLOW_CONTROL:TRUE
This is a dynamic setting which means the instance does not have
to be restarted to pick up the registry
variable setting. With flow control enabled, global snapshots
will not consume a large number of FCM buffers. |
| Workaround |
not known / see Local fix |
| Timestamps |
Date - problem reported :
Date - problem closed :
Date - last modified :
| 18.04.2011
27.06.2011
27.06.2011 |
| Problem solved at the following versions (IBM BugInfos) |
9.5.FP8,
9.5.FP8
|
| Problem solved according to the fixlist(s) of the following version(s) |
| 9.5.0.8
|
 |
