G06.27 ReadMe

o NonStop Kernel (T9050)
Solution 10-050706-9478
When the $ZZKRN policy for DCT size is changed from SMALL to
MEDIUM, on system load the policy change is not reflected in
"PEEK /CPU nn/ POOL" info.
Recovery
Issue an "alter subsys $ZZKRN, DCT medium" command in SCF after
a system load is performed.
Workaround
Issue an "alter subsys $ZZKRN, DCT medium" command in SCF after
a system load is performed.
o OSS NAME SERVER (T8621)
Solution 10-050219-4847
The Name Server process drives the CPU busy for a long time
(potentially hours) and Measure data shows a large number of
messages from Name Server to DP2 (possibly millions). This occurs
under the following conditions:
a) The number of files in a directory of a fileset is more than
the configured LINKCACHE limit of the associated Name Server.
AND
b) Name Server receives name lookups (for OSS name from Guardian
name) for every file in the directory ( for example, Measure
with OSS journaling ON).
Recovery
The problem eventually goes away (when Measure runs out of files
to do journal entries for.)
Workaround
a) Increase the value of LINKCACHE of Name Server through SCF to
at least the number of files in the largest directory of the
filesets associated with the Name Server.
OR
b) Split the large directories into smaller directories.
o OSS NAME SERVER (T8621)
Solution 10-050624-9126
Name Server process' CPU consumption is high and the system
performance (for example, OLTP applications running in the CPU
where NS process is running) is affected.
The problem would occur under all of the following conditions:
- When there is a large number of files in a directory of a
fileset. (For example, 35000 files in a directory.)
- A FIND command is given on a file in the directory.
Recovery
None. Once the FIND command completes, the performance goes back
to normal.
Workaround
Rearrange the directory structure and keep the number of files in
Page
3
of
5