[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 20000216: [Fwd: LDM version 5.0.9] (fwd)


===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================

---------- Forwarded message ----------
Date: Fri, 18 Feb 2000 18:10:44 -0500
From: Jeff Ator <address@hidden>
To: Anne Wilson <address@hidden>
Subject: Re: 20000216: [Fwd: LDM version 5.0.9]

Anne Wilson wrote:

> Jeff,
>
> We have a hypothesis.  It is the case that the
> products you are receiving are relatively small.
> The LDM assumes a default product size of 2048 and
> calculates the number of product slots by
> <queueSize>/2048. Your products tend to be
> signficantly smaller.  Thus, it is possible that
> your queue could run out of slots.
>
> I understand that everything was working under
> IRIX 5.0.5.  But, since you allocated 75Mb perhaps
> that gave you enough slots that the queue would
> not need to grow.  Please look at the log entries
> for this configuration and see what the queue
> usage is there.
>

Actually, even the 5.0.5 queue has grown some, albeit over a long period
of time, under IRIX64.  I apologize if I didn't make this clear in our
earlier communications.  It was initially allocated with "ldmadmin
mkqueue -s 75000000" on 7/29/1999, and is currently at around 84.3Mb.
So it has grown some, but not very much, and all this time we have been
running pqexpire from cron with "-a 12" to always keep the most recent
12 hours of data in the queue.  A recent output from one of these
pqexpire runs shows that the queue usage is good, around 75Mb:

Feb 18 14:51:04 pqexpire[4932]: Starting Up
Feb 18 14:56:24 pqexpire[4932]: Exiting
Feb 18 14:56:24 pqexpire[4932]: > Up since:      20000218145104.783
Feb 18 14:56:24 pqexpire[4932]: > Queue usage (bytes):75005952
Feb 18 14:56:24 pqexpire[4932]: >          (nregions):  128589
Feb 18 14:56:24 pqexpire[4932]: > nbytes recycle:      4068520 (
4040.195 kb/hr)
Feb 18 14:56:24 pqexpire[4932]: > nprods deleted:         8158 (
8295.634 per hour)
Feb 18 14:56:24 pqexpire[4932]: > First deleted: 20000218015200.014
Feb 18 14:56:24 pqexpire[4932]: > Last  deleted: 20000218025100.286


>
> Also, we would like you to test 5.0.9 on IRIX
> using identical parameters to 5.0.5.  That is,
> start with a 75Mb queue.  Please let us know the
> results, especially the queue usage.

OK, I began this test at around 1630z this morning.  Specifically, I
created new ldm.pq files at 75Mb under 5.0.9 on our AIX4.3 and IRIX64
machines and have been forking a copy of our regular data feed to each
(via the 5.0.9 version of pqing) since 1632z and 1649z, respectively.
On the AIX4.3 box, I began seeing "Deleting oldest to get a queue slot"
messages at 2032z (after exactly 4 hours); whereas, on the IRIX64 box, I
began seeing "Growing index at 16384" messages at 2058z (after 4 hours,
9 minutes).  So the "filling up" of the queue appears comparable on both
systems under 5.0.9.  It will be late tonight before 12 hours of the
test have elapsed and I can get a good feel for how big the IRIX64 5.0.9
queue grows to hold that much data.  At that point, we could compare its
size with that of the IRIX64 5.0.5 queue and see if it has also grown to
around 85Mb.  If so, then it may be that the 5.0.5 and 5.0.9 versions
are comparable in how big a queue it takes to hold 12 hours of data, and
we can then shift our focus to the peculiarities of mmap under AIX.
Unfortunately, I will be away from the office (and probably not able to
check on the test results!) until next Thursday, so I will email you
those results at that time.  The test will have run for almost a week at
that point, so we should have some fairly reliable results.  Anyway, for
now, here are the pqexpire results (showing queue usage statistics) for
the past few hours (i.e. since the start of the test) on the:

AIX4.3 (5.0.9)

Feb 18 16:51:01 nco3n03 pqexpire[28758]: Starting Up
Feb 18 16:51:06 nco3n03 pqexpire[28758]: Exiting
Feb 18 16:51:06 nco3n03 pqexpire[28758]: > Up since:
20000218165101.196
Feb 18 16:51:06 nco3n03 pqexpire[28758]: > Queue usage (bytes): 1341360
Feb 18 16:51:06 nco3n03 pqexpire[28758]: >          (nregions):    3907
Feb 18 16:51:06 nco3n03 pqexpire[28758]: > nprods deleted 0
Feb 18 17:51:01 nco3n03 pqexpire[19050]: Starting Up
Feb 18 17:51:23 nco3n03 pqexpire[19050]: Exiting
Feb 18 17:51:23 nco3n03 pqexpire[19050]: > Up since:
20000218175101.317
Feb 18 17:51:23 nco3n03 pqexpire[19050]: > Queue usage (bytes): 5542640
Feb 18 17:51:23 nco3n03 pqexpire[19050]: >          (nregions):   12558
Feb 18 17:51:23 nco3n03 pqexpire[19050]: > nprods deleted 0
Feb 18 18:51:00 nco3n03 pqexpire[28656]: Starting Up
Feb 18 18:51:47 nco3n03 pqexpire[28656]: Exiting
Feb 18 18:51:47 nco3n03 pqexpire[28656]: > Up since:
20000218185100.566
Feb 18 18:51:47 nco3n03 pqexpire[28656]: > Queue usage (bytes):10600544
Feb 18 18:51:47 nco3n03 pqexpire[28656]: >          (nregions):   22034
Feb 18 18:51:47 nco3n03 pqexpire[28656]: > nprods deleted 0
Feb 18 19:51:00 nco3n03 pqexpire[28508]: Starting Up
Feb 18 19:51:59 nco3n03 pqexpire[28508]: Exiting
Feb 18 19:51:59 nco3n03 pqexpire[28508]: > Up since:
20000218195100.187
Feb 18 19:51:59 nco3n03 pqexpire[28508]: > Queue usage (bytes):15133528
Feb 18 19:51:59 nco3n03 pqexpire[28508]: >          (nregions):   31078
Feb 18 19:51:59 nco3n03 pqexpire[28508]: > nprods deleted 0
Feb 18 20:51:00 nco3n03 pqexpire[18072]: Starting Up
Feb 18 20:52:06 nco3n03 pqexpire[18072]: Exiting
Feb 18 20:52:06 nco3n03 pqexpire[18072]: > Up since:
20000218205100.977
Feb 18 20:52:06 nco3n03 pqexpire[18072]: > Queue usage (bytes):18254888
Feb 18 20:52:06 nco3n03 pqexpire[18072]: >          (nregions):   36692
Feb 18 20:52:06 nco3n03 pqexpire[18072]: > nprods deleted 0
Feb 18 21:51:00 nco3n03 pqexpire[30516]: Starting Up
Feb 18 21:52:07 nco3n03 pqexpire[30516]: Exiting
Feb 18 21:52:07 nco3n03 pqexpire[30516]: > Up since:
20000218215100.355
Feb 18 21:52:07 nco3n03 pqexpire[30516]: > Queue usage (bytes):18719608
Feb 18 21:52:07 nco3n03 pqexpire[30516]: >          (nregions):   36692
Feb 18 21:52:07 nco3n03 pqexpire[30516]: > nprods deleted 0
Feb 18 22:51:01 nco3n03 pqexpire[25094]: Starting Up
Feb 18 22:52:07 nco3n03 pqexpire[25094]: Exiting
Feb 18 22:52:07 nco3n03 pqexpire[25094]: > Up since:
20000218225101.028
Feb 18 22:52:07 nco3n03 pqexpire[25094]: > Queue usage (bytes):18750328
Feb 18 22:52:07 nco3n03 pqexpire[25094]: >          (nregions):   36692
Feb 18 22:52:07 nco3n03 pqexpire[25094]: > nprods deleted 0

and IRIX64 (5.0.9)

Feb 18 17:51:04 5Q:sgi98 pqexpire[19576]: Starting Up
Feb 18 17:51:06 5Q:sgi98 pqexpire[19576]: Exiting
Feb 18 17:51:06 5Q:sgi98 pqexpire[19576]: > Up since:
20000218175104.455
Feb 18 17:51:06 5Q:sgi98 pqexpire[19576]: > Queue usage (bytes): 4222752

Feb 18 17:51:06 5Q:sgi98 pqexpire[19576]: >          (nregions):    8814

Feb 18 17:51:06 5Q:sgi98 pqexpire[19576]: > nprods deleted 0
Feb 18 18:51:05 5Q:sgi98 pqexpire[22089]: Starting Up
Feb 18 18:51:20 5Q:sgi98 pqexpire[22089]: Exiting
Feb 18 18:51:20 5Q:sgi98 pqexpire[22089]: > Up since:
20000218185105.247
Feb 18 18:51:20 5Q:sgi98 pqexpire[22089]: > Queue usage (bytes): 9175760

Feb 18 18:51:20 5Q:sgi98 pqexpire[22089]: >          (nregions):   18377

Feb 18 18:51:20 5Q:sgi98 pqexpire[22089]: > nprods deleted 0
Feb 18 19:51:04 5Q:sgi98 pqexpire[2105]: Starting Up
Feb 18 19:51:16 5Q:sgi98 pqexpire[2105]: Exiting
Feb 18 19:51:16 5Q:sgi98 pqexpire[2105]: > Up since:
20000218195104.958
Feb 18 19:51:16 5Q:sgi98 pqexpire[2105]: > Queue usage (bytes):13677872
Feb 18 19:51:16 5Q:sgi98 pqexpire[2105]: >          (nregions):   27255
Feb 18 19:51:16 5Q:sgi98 pqexpire[2105]: > nprods deleted 0
Feb 18 20:51:04 5Q:sgi98 pqexpire[11247]: Starting Up
Feb 18 20:51:33 5Q:sgi98 pqexpire[11247]: Exiting
Feb 18 20:51:33 5Q:sgi98 pqexpire[11247]: > Up since:
20000218205104.436
Feb 18 20:51:33 5Q:sgi98 pqexpire[11247]: > Queue usage (bytes):17734192

Feb 18 20:51:33 5Q:sgi98 pqexpire[11247]: >          (nregions):   35737

Feb 18 20:51:33 5Q:sgi98 pqexpire[11247]: > nprods deleted 0
Feb 18 21:51:05 5Q:sgi98 pqexpire[13854]: Starting Up
Feb 18 21:51:25 5Q:sgi98 pqexpire[13854]: Exiting
Feb 18 21:51:25 5Q:sgi98 pqexpire[13854]: > Up since:
20000218215105.125
Feb 18 21:51:25 5Q:sgi98 pqexpire[13854]: > Queue usage (bytes):22075696

Feb 18 21:51:25 5Q:sgi98 pqexpire[13854]: >          (nregions):   44431

Feb 18 21:51:25 5Q:sgi98 pqexpire[13854]: > nprods deleted 0
Feb 18 22:51:04 5Q:sgi98 pqexpire[12959]: Starting Up
Feb 18 22:51:28 5Q:sgi98 pqexpire[12959]: Exiting
Feb 18 22:51:28 5Q:sgi98 pqexpire[12959]: > Up since:
20000218225104.659
Feb 18 22:51:28 5Q:sgi98 pqexpire[12959]: > Queue usage (bytes):26171472

Feb 18 22:51:28 5Q:sgi98 pqexpire[12959]: >          (nregions):   52523

Feb 18 22:51:28 5Q:sgi98 pqexpire[12959]: > nprods deleted 0

Again, I'll email you again next Thursday, and we'll see what the queue
usage statistics are at that time.


>
>
> Regardless of the outcome of the above test, you
> can use the -S option in pqcreate to specify the
> number of slots you want for your queue.  Give
> this a try, but, if possible, please do the above
> tests first so we can confirm our suspicions.
>
> By the way, we are aware that the -s option to
> pqcreate is not working.  Our apologies for that.
>
> Also, I understand that it is often easier to talk
> rather than send email.  However, email is
> important to us because we use it to track our
> efforts, interact with each other, and populate
> our database.  So, please send email in the
> future.  Thank you!

Hey, no problem!  I appreciate all of your help!

Thanks again,
-Jeff


--
__________________________________

Jeffrey B. Ator

National Centers for
   Environmental Prediction
Central Operations
5200 Auth Road, Rm. 307
Camp Springs, MD  20746

Phone:  (301) 763-8000, Ext. 7104
Fax:    (301) 763-8381

E-Mail:  address@hidden

__________________________________