 
| Subject: | |
| From: | |
| Reply To: | |
| Date: | Thu, 13 Mar 2003 22:12:43 GMT |
| Content-Type: | text/plain |
| Parts/Attachments: |
|
|
Hello All,
RE: MPE/iX 7 and sendmail
======================================================================
Buf Free Total Malloc Free
Size Bufs Bufs delta Calls Calls delta bc depth pend cb
1 0 0 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0 0 0
4 0 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0 0 0 0 0 0
16 0 0 0 0 0 0 0 0 0 0
32 0 0 0 0 0 0 0 0 0 0
64 127 128 1 107 106 1 0 0 0 0
128 48 64 16 301 285 16 0 0 0 0
256 32 32 0 204 204 0 0 0 0 0
512 11 16 5 114 109 5 0 0 0 0
1024 8 8 0 126 126 0 0 0 0 0
2048 3 4 1 3 2 1 0 0 0 0
4096 0 0 0 0 0 0 0 0 0 0
8192 0 0 0 0 0 0 0 0 0 0
Statistics:
Memory freeze daemon waken up: 0
Memory thaw daemon waken up: 0
Freeze segment on user stack: 1
Couldn't freeze segment on user stack: 0
Frozen free segments: 19 Unfrozen free segments:487
steal_segment succeeded/called: 0 / 0
steal_bigger_buffer succeeded/called: 0 / 0
Total memory frozen in buckets (bytes): 49152
Total memory available in buckets (bytes): 42432
--------------------------------------------------
I found this very interesting. I'm curious about
the display and what it is actually showing me.
Here's the one from my 989-650 running 6.5pp3 with
3.7 Gig of memory. What are the "buckets"
referring to, and why do I have only a total of
90k there? Is there a basic problem that I've
overlooked in various upgrades over the last three
years? Could this be an indicator of one of my
issues with poor performance? Any light you can
shed is greatly appreciated.
--------------------------------------------------
I really was only interested in seeing the "Total memory available in
buckets" at the time the error was seen:
> :sh.hpbin.sys
> $ echo | cat
> no pipes available: Not enough memory space
> Internal error: j_close: no processes
As you can see the 'streams' implementation on the 3k allocates a
small amount of memory:
:calc 49152+42432
91584, $165C0, %262700
1048576
:calc (49152+42432)/1024
89, $59, %131
...1048576 Bytes or 89Kb of memory as its managed buffer area.
Applications which make calls to the streams procedures will
"allocate" and "free" buffers in this managed memory buffer area.
The size of this buffer area is hardcoded in the streams code and is
not configurable. We have not had reported (KP) Known Problems
associated with running out of this buffer space, but the above pipes
error would lead me to suspect it was the case.
--------------------------------------------------
>> I found this very interesting. I'm curious
>> about the display and what it is actually
>> showing me.
--------------------------------------------------
It is displaying 89Kb of memory managed by the streams subsystem
for pipe calls and any subsystem making calls to streams procedures
to "allocate" and "free" buffers.
At this point I am not aware of any subsystems or products making
calls to streams procedures other than pipe calls, but there
might be some products available on the market which make use of
this subsystem. AFAIK HP products are only making use of the
pipe "|" function.
--------------------------------------------------
>> Here's the one from my 989-650 running 6.5pp3
>> with 3.7 Gig of memory. What are the "buckets"
>> referring to, and why do I have only a total of
>> 90k there?
--------------------------------------------------
The buckets are the managed units that the streams buffer manager
works with. I expect each managed buff size is a bucket, but I am
going to pass on searching the source to verify this. The 89Kb of
memory managed by the streams subsystem is hardcoded.
--------------------------------------------------
>> Is there a basic problem that I've overlooked
>> in various upgrades over the last three years?
>> Could this be an indicator of one of my issues
>> with poor performance? Any light you can shed
>> is greatly appreciated.
--------------------------------------------------
No (KB) Known Problems are associated with the streams buffer manager
in a full condition as it requires no configuration and is hardcoded
at the same size for all MPE systems. The streams buffer manager
allocates it's memory when it is started for the first time... When
the streams subsystem is started (typically at the first pipe "|"
call) you will see console messages:
--------------------------------------------------
MPS: installed device clone as major 1
MPS: installed device sad as major 121
MPS: installed device pipe as major 137
MPS: installed module sw2proc
MPS: installed module sc
MPS: installed module pipemod
MPS: installed device log as major 122
MPS: installed module timod
MPS: installed module tirdwr
MPS: installed device loop as major 124
Starting Streams/iX memory freeze process [pin 101]
Starting Streams/iX memory release process [pin 102]
Starting Streams/iX scheduler [pin 103]
--------------------------------------------------
I am suspect that the problem where the error "Not enough memory
space" which was seen is most likely caused by an application which is
allocating and not freeing memory buffers in streams. It could also
be due to a corrupt pointers resulting in leaked buffers... Finally
it could be caused by "pipe read/write hangs" fixed as per SR
8606134134.
I hope this helps.
Regards,
James Hofmeister
Hewlett Packard - Global Solutions Engineering (WTEC)
P.S. My Ideals are my own, not necessarily my employers.
Hmmmm... "The great tragedy of science - the slaying of a beautiful hypothesis by an ugly fact." - T. H. Huxley
________________________________________________________________
Sign Up for Juno Platinum Internet Access Today
Only $9.95 per month!
Visit www.juno.com
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
|
|
