| Author |
Message
|
| sebastia |
 Posted: Fri Jan 30, 2009 6:09 am Post subject: how to calculate "-LS" from "-LP" |
 |
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Hi.
Lets supose I need a 25 GB log file on AIX, MQ version 6.
If I set "-lf 65535" then each log file will have 268 MB.
Then, my calculations lead me to have "-lp 100".
So far, so good.
Now I have to decide on "-ls", secondary files number.
As each file will be 269 MB, it is not the same to have 10 or to have 50.
How do I calculate this number ?
Seb. |
|
| Back to top |
|
 |
| mqjeff |
| Posted: Fri Jan 30, 2009 6:22 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
The total size of all the log should be no smaller than necessary and no bigger than you can stand.
There are two schools of thought on the the ratio of lp to ls. One says "lp should be all I ever need, and ls is just there for emergency". The other says "lp should be there for what I typically use, and ls is there for high volume cases".
If you are of the first opinion, then you would make -lp large enough to fill up all but say 1 of those 25GB, and leave -ls to use the rest. If you are of the second opinion, you would likely split the 25GB more equally. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Fri Jan 30, 2009 6:41 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| The total size of all the log should be no smaller than necessary and no bigger than you can stand. |
More concretely, of the three valuable computer resources (RAM, processors, disk space), disk space is usually the lease expensive.
For circular logging, you must have more log segments than the longest unit of work can span. If you intend to archive your log segments, then you must have enough other log segments to allow the archive process to complete before you run out of log segments.
Primary log segments are formatted when the queue manager is created; secondaries are formatted and used when needed - and where needed.
As Jeff suggested, you need to calculate log size based on both anticipated (lp), unanticipated (ls) workload, plus some percentage fudge-factor (unanticipated unanticipated workload).
The WMQ System Admin manual has a section on calculating the logs. It is a good beginning. Your application developers must contribute to this process. They should know message sizes, message rates, UofW duration.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| sebastia |
| Posted: Fri Jan 30, 2009 7:05 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Jeff - I would like to choose between the two options you specify,
but I see (saw) no diference between Primary and Secondary.
Without more detail, I would choose the FIRST school of thinking ( )
Bruce - you provided the light I needed to discriminate
LP = pre-formatted disk space, used for anticipated workload
LS = not-formatted disk space, used for un-anticipated traffic
As the 25 GB are all "anticipated", I shall get another 10% for the
un-anticipated, so 2 GB more in "secondary" space,
if such big space (27GB) is available, of course
I know it uses to happen than when my calculations require a lot of disk,
probably not available,
customer comes with "lets log only 3 days instead of 5",
then making the existing disk big enough.
Thanks you all. |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Sun Feb 01, 2009 11:08 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
| Quote: |
| "lets log only 3 days instead of 5" |
Are you using circular or linear logs?
The purpose of the logs is to enable the qmgr to restart in a consistent state. At restart, IF NECESSARY, the qmgr will replay the logs. If the qmgr shuts down normally (writes a closing checkpoint), it will restart pretty quickly. The only log segments that are needed for restart are those that contain active UofW's at shutdown - which under normal circumstances should be zero.
For what (other) purpose is your customer keeping 3 days or 5 days of logs?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| sebastia |
| Posted: Wed Feb 04, 2009 7:56 am Post subject: |
|
|
Grand Master
Joined: 07 Oct 2004
Posts: 1003
|
Hi, Bruce - sorry for the delay ... I am away from home right now.
1) log type
All my customers use Circular logs (eitherz/Linux, AIX and Windows),
even T.Rob told me he is very used (or his customers) tu use linear.
Still have not found the need for "media recovery".
Maybe you (or any other listener) can tell me in which situation
this may be needed ?
When a specific message is "deleted" from a queue and
we want to recover it ?
Is this the meaning of "media recovery" ?
b) 3 or 5 days of data in Log.
This is the kind of question I do not present to my customers.
I ask "how many days do you want to be able to recover",
and they did answer 3 or 5.
No application details are never given to me, once I say
"we need 30GB of raid disk" and they say "here you are".
Cheers. Sebastian. |
|
| Back to top |
|
|
| mqjeff |
| Posted: Wed Feb 04, 2009 7:59 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
| You need media recovery when you have messages sitting on queues for more than a short period of time, and need to be able to recover the messages without recreating them from sending applications. |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed Feb 04, 2009 8:05 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
I make LS 10% of LP. Since ther eis no good answer for this question, I just made that decision and we use it across the board. Right or wrong, at least we are consistent.
Linear Log question:
If you do media recovery, its going to recreate the object and all the messages in the queue at the point in time that the image was last taken, plus the logs since then. There is no way to tell the QM to play the messages only up to 12:34:01.1452. If the media image was taken at 12:00:00.000, and you are at 01:00:00.00, and you do media recovery for the q, it will look like it did at 01:00:00.000, right?
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| mqjeff |
| Posted: Wed Feb 04, 2009 8:11 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008
Posts: 17447
|
| PeterPotkay wrote: |
| If the media image was taken at 12:00:00.000, and you are at 01:00:00.00, and you do media recovery for the q, it will look like it did at 01:00:00.000, right? |
Assuming you have log files from the current moment, yes. If, perhaps, you have log files from 12:00:00.000 and you use those instead... |
|
| Back to top |
|
|
| PeterPotkay |
| Posted: Wed Feb 04, 2009 8:24 am Post subject: |
|
|
Poobah
Joined: 15 May 2001
Posts: 7722
|
So I think this is a common misconception out there about Linear Logs. Some people think that they can play the linear logs to make the queue look like it did at a certain point, so they can get messages from the q that existed at 12:34:01.1452. Not so (without a 3rd party tool like Cressida perhaps), they will get the q as it looked like right up the the latest log point.
_________________
Peter Potkay
Keep Calm and MQ On |
|
| Back to top |
|
|
| exerk |
| Posted: Thu Feb 05, 2009 12:03 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
| mqjeff wrote: |
| You need media recovery when you have messages sitting on queues for more than a short period of time, and need to be able to recover the messages without recreating them from sending applications. |
I understood the media recovery aspect of linear logging was to recover objects marked as damaged, rather than replaying messages. Could you please elaborate as to the mechanics of using it to recover messages?
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| Vitor |
| Posted: Thu Feb 05, 2009 1:05 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| exerk wrote: |
| Could you please elaborate as to the mechanics of using it to recover messages? |
Restart the queue manager.
It will to a forward recovery to the point-in-time for the logs.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| exerk |
| Posted: Thu Feb 05, 2009 1:16 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
Sorry, badly formed question on my part. I was trying to get a reconciliation between PeterPotkay's post (which is as I understand object recovery to be) and that of mqjeff's post as the implication in the post was that it is possible to recover specific messages from a point in time.
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| Vitor |
| Posted: Thu Feb 05, 2009 1:25 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
A linear log will recover a damaged object and the messages on it automatically at start up.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| exerk |
| Posted: Thu Feb 05, 2009 1:32 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
To be yet more specific, I read the post as an implication that using the rcrmqobj command would be used in the process.
It's been a long week for the first one back at the grind after Christmas!
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
|
|
