FCM
Troubleshooting
For the
most part, a DBA will not know that FCM really exists. The DB2 instance has
been pre-optimized for a variable load, with satisfying default parameters.
However, FCM buffers are not self-tuned and need to be observed. Let's look at
one of the frequent problems with FCM buffers:
Reactive
database administrator
Early
failure occurrence from the database message log (db2dump.log):
Considering
log file and error messages inside, there is an indication of a serious problem
with the database resource alloc_buffer. DBM has to react to that
error condition either by closing applications or in extreme cases by closing
the database.
Failure
endpoint from the database message log, where DBM is closing all of the database
connections:
base_sys_utilities stopdbm Probe:911
Database manager is stopped.
2003-08-26-17.39.55.191185 Instance:artist Node:000
PID:27123(db2stop2) Appid:none
base_sys_utilities stopdbm Probe:911
fast_comm_manager sqlkf_process_goodbye Probe:40
Node 0 has closed its connection.
2003-08-26-17.39.51.787015 Instance:artist Node:002
PID:1510(db2fcmdm 4) Appid:none
fast_comm_manager sqlkf_process_goodbye Probe:40
Node 2 has closed its connection.
2003-08-26-17.39.51.787015 Instance:artist Node:001
PID:1510(db2fcmdm 4) Appid:none
fast_comm_manager sqlkf_process_goodbye Probe:40
Node 1 has closed its connection.
Listing
system message log, showing many FCM errors:
# cat /var/adm/messages
Aug 26 17:39:53 ARTIST01 DB2[1506]: [ID 212799 user.error]
DB2 (artist.002) fast_comm_manager sqlkf_process_goodbye reports: probe id 40 with error 0 and alert num 0
Aug 26 17:39:53 ARTIST01 DB2[1506]: [ID 482771 user.error]
extra symptom string provided: Node 1 has closed its connection.
Aug 26 17:45:14 ARTIST01 DB2[6731]: [ID 860831 user.error]
DB2 (artist.000) fast_comm_manager sqlkfMlnPreConnect reports: probe id 31 with error 2055 and alert num 0
Aug 26 17:45:14 ARTIST01 DB2[6731]: [ID 139117 user.error]
extra symptom string provided: Failed to send hsh msg to node =
Aug 26 17:45:14 ARTIST01 DB2[6731]: [ID 702911 user.error] data: 00000001 ....
Aug 26 17:45:14 ARTIST01 DB2[6731]: [ID 860831 user.error]
DB2 (artist.000) fast_comm_manager sqlkfMlnPreConnect reports: probe id 31 with error 2055 and alert num 0
Aug 26 17:45:14 ARTIST01 DB2[6731]: [ID 139117 user.error]
extra symptom string provided: Failed to send hsh msg to node =
Aug 26 17:45:14 ARTIST01 DB2[6731]: [ID 702911 user.error] data: 00000002 ....
Listing application
error message:
Error: QueryEngine encountered error: Call SQLFetch got SQL_ERROR
[57011:-6040: on HSTMT] [IBM][CLI Driver][DB2/SUN] SQL6040C No available FCM-Buffer. SQLSTATE=57011
Explaining
DB2 error SQL6040:
db2 => ? sql6040
SQL6040C No FCM buffers are available.
Explanation: No FCM buffer is available.
The statement cannot be processed.
User Response: Try the request again.
If the error persists, increase the number of FCM buffers (fcm_num_buffers)
specified in the database manager configuration file, then try the request again.
sqlcode: -6040
sqlstate: 57011
The reactive
DBA has now all of the information needed to know how to fix the problem, (and
he should hope that this would not happen again).