Here is a line from the log file of one of the replication servers in
my system. I'm new to System10 replication server. The system is HP-UX.

ERROR #7031 SQT(103:1 LNa_P.position) -m/memseg.c(702)
    Cannot expand memory since '64' memory maps already exist.

At the same time one SQT and one DIST thread are down too.

Any help in diagnosing this problem will be appreciated.

Surya.

PS: You can reply directly to me if you want.

Surya,

admin statistics, mem will show you exactly how memory was used.
typically it shows that a lot of memory is available, but the repserver
ran out of a certain segment size. i will not go into all the details of
how memory management works, but it is designed to prevent fragmentation
(which is important for a long running server), rather than optimize
memory utilization. therefore, it allocates memory in chunks of a certain
size (called segments) and reuses these segments. sometimes it runs out
of chunks of a certain size and you get the message you encountered.
often, rebooting the replication server is all it takes. if the problem
persists or if admin statistics, mem shows that most of the memory is in
use, i would add more memory.

you can expand the memory used by changing the
value for the "memory_max" parameter in the rs_config table in RSSD.
See page 2-13 of the Replication Server Administrator's guide. The
default value is 3Meg (i think). Try setting it to 5.

also, sometimes memory problems are caused by very large
transactions. SQT tries to cache them and runs out of memory.
if you have large transactions, i would also expand the sqt_max_cache_size.

i would also debug the system a little more.
try admin who,sqt to look for orphaned transactions; admin who,sqm to
look at your storage.
troubleshooting guide may give you some more ideas.

hope this helps. alex
Replication Server consultant

--

------------------------------------------------------------------------------
Alex Gorelik                                 Acta
Systems Architect       Practical Solutions for the Distributed Enterprise

Repserver's memory management uses segments as the lowest level of internal
allocation these segments can have sizes varying 256 bytes to 64KB.
Segments of the same size are grouped together in "maps".
When the server requests memory from the operating system it does so by asking
for a chunk of memory which it splits up into segments. For any chunk the
segments are all the same size. These chunks are identified by the memory
maps.
At startup the server requests one chunk of each size of segments (256, 1K
4K, 16K, 64K) so 5 memory maps are used. During runtime if it runs out of
a particular size of segment it allocates a new chunk using another memory
map. Memory maps are never released so it is possible to have a large number
of unused segments and still need to allocate a new chunk of a different size
segments.
The server only has 64 of these map structures available so once they are
used up any requests for new chunk of memory will get the above error.
The immediate solution is to restart the repserver as this will reset the
maps.
The size of the chunk allocated to each map is based on the servers memory
config. multplied by a scaling factor which I can't remember. The 5 chunks
allocated at startup are bigger than the ones allocated later and the total
which can be allocated is considerably bigger than the config. value.

I remember seeing a very good write up of this (a much better explanation
then mine) and other repserver questions, but I have lost my copy. If this
doc. is available on line anywhere I would be grateful for a pointer to it.

Cheers,

Pete.