| Author |
Message
|
| atommail |
 Posted: Tue Nov 17, 2009 4:53 am Post subject: bad practise or not? |
 |
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
Hello
I have to design request-reply scenario in websphere message broker toolkit.
I read that set CorrelId is bad practise, but could anybody tell my why?
or could anybody tell alternative from my scenario?
SystemY response in asynchronic way.
Flow in systemY_adapter_request:
MQInput-->ComputeNode-->HTTPRequest-->ComputeNode(storeState: set CorrelId using myId, store MqReply)-->MQOutput(storeState)
Flow in systemY_adpater_response
HTTPInput-->ComputeNode (setCorrelId using myId from HTTPInput)-->MQGet(getParameters by CorrelId ex. replyToQueue)-CopmuteNode -->MQReply
systemX_adapter flow put in systemY_adapter_request and get response from systemY_adpater_response flow --> this is synchronic scenario
I've already developed this and it is working, but I don't know, can I use this at production, Is it a bad practise or not? I don't have experience, because it is my first project  please help me |
|
| Back to top |
|
 |
| zpat |
| Posted: Tue Nov 17, 2009 5:11 am Post subject: |
|
|
Jedi Council
Joined: 19 May 2001
Posts: 5866
Location: UK
|
In general terms it is a normal design for a MQ request message to be processed by a server application which copies the msgid of the request message to the correlid of the response message.
The requestor then gets the response to their own request matching by correlid on the reply queue, and only two queues need to be defined (requests and responses) along with remote versions as needed.
One point is that response messages should have an expiry value set so that if they are not picked up then they expire. The design should allow many requestors to use the same set of queues at the same time. |
|
| Back to top |
|
|
| atommail |
| Posted: Tue Nov 17, 2009 5:47 am Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
zpat,
Ok, but this is common scenario using msgId and correlId,
How can I aply this to my scenario, I think I can't ?
Idon't have queue nodes but HTTP Request and asynchronic response from system Y to response flow. Response from System Y: system Y call webservice in systemY_adapter, system Y doesn't put response to queue, it call webservice, so I don't have msqId. |
|
| Back to top |
|
|
| atommail |
| Posted: Tue Nov 17, 2009 5:58 am Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
I have orderId and I use orderId instead of msqId
SET OutputRoot.MQMD.CorrelId = orderId;
Is it correct?
I don't have another idea. |
|
| Back to top |
|
|
| Vitor |
| Posted: Tue Nov 17, 2009 7:57 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| atommail wrote: |
| I don't have msqId. |
You have the original msgid from the original request via MQInput.
To clarify your original question:
| atommail wrote: |
| I read that set CorrelId is bad practise |
setting correlId is not a bad practice. What is bad practice is using non-system ids (i.e. not msgId or correlId) because you can easily hit problems, mostly around non-uniquness. In your example, consider what would happen if the same order was changed or processed at the same time (given that you're using order id). Clearly you can design business or other processes to get round that but then you're reliant on 2 things working correctly...
Fewest moving parts, best practice. 
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| atommail |
| Posted: Tue Nov 17, 2009 9:53 am Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
thanks for reply,
| Quote: |
| You have the original msgid from the original request via MQInput. |
Yes, I have, but I have HTTPRequest Node in response flow and I don't have msqId here, so still I have to take my scenario.
My orderId is unique,
consist of: customerId and CURRENT_TIMESTAMP AS CHARACTER FORMAT 'yyyyMMdd.HHmmss
Next I convert it to BLOB, so I think there is no way that i will have two messages in my storage_state_queue with the same CorrelId?
My question is: will be something wrong if somwhere in another flow, another queue be message with the same CorrelId (QueueManager set exactly the same what I set in my flow)? |
|
| Back to top |
|
|
| Vitor |
| Posted: Tue Nov 17, 2009 9:59 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| atommail wrote: |
| Yes, I have, but I have HTTPRequest Node in response flow and I don't have msqId here |
Doesn't mean you couldn't use it.
| atommail wrote: |
My orderId is unique,
consist of: customerId and CURRENT_TIMESTAMP AS CHARACTER FORMAT 'yyyyMMdd.HHmmss
Next I convert it to BLOB, so I think there is no way that i will have two messages in my storage_state_queue with the same CorrelId? |
And relies on one order per customer per second.
| atommail wrote: |
| My question is: will be something wrong if somwhere in another flow, another queue be message with the same CorrelId (QueueManager set exactly the same what I set in my flow)? |
Yes.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| atommail |
| Posted: Tue Nov 17, 2009 10:54 am Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
thanks a lot for your respond
So, I have to change ma scenario, Could you help me?
| Vitor wrote: |
| atommail wrote: |
| Yes, I have, but I have HTTPRequest Node in response flow and I don't have msqId here |
Doesn't mean you couldn't use it.
l
|
I did mistake, I should write: "Yes, I have, but I have HTTPInput"
Could you tell me how I could use msgId?
How can I correlate request and replay I can't imagine this because this is two flows and when Response flow get Response call Response flow doesn't know anything about Request flow |
|
| Back to top |
|
|
| Vitor |
| Posted: Tue Nov 17, 2009 11:22 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| atommail wrote: |
| Could you tell me how I could use msgId? |
The same way you're using your orderid - the only point I'm making here is that it's a "better" id than the one you're generating.
| atommail wrote: |
| How can I correlate request and replay |
In your first post you said you had this working - stick to that.
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| atommail |
| Posted: Tue Nov 17, 2009 11:39 pm Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
I am a bit confused now ...
So, I have to make better id , and then will be ok? |
|
| Back to top |
|
|
| AkankshA |
| Posted: Wed Nov 18, 2009 12:42 am Post subject: |
|
|
Grand Master
Joined: 12 Jan 2006
Posts: 1494
Location: Singapore
|
| Quote: |
I am a bit confused now ...
So, I have to make better id , and then will be ok?
|
what's your confusion... spell it out..
you need not make a better id.. just use the other id.. which is considered better.. msg id
and correlation mechanism is same as you are using...
_________________
Cheers |
|
| Back to top |
|
|
| atommail |
| Posted: Wed Nov 18, 2009 1:15 am Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
My response flow:
HTTPInput -->ComputeNode (I set CorrelId here)-->MQGet -->other things
In Compute node I don't have msgId to use it, so I use my orderId
Everybody tell me that I should use msgId and I shouldn't use my orderId to set CorrelId, I know that, I should, but I can't, I don't have msgId.
One way, I see is make my orderId better, mean uniqness,
my idea myabe add seconds when I build it? |
|
| Back to top |
|
|
| AkankshA |
| Posted: Wed Nov 18, 2009 1:29 am Post subject: |
|
|
Grand Master
Joined: 12 Jan 2006
Posts: 1494
Location: Singapore
|
ohhh.. ok ok
so you mean to say that you create a new unique id in request flow and send it as msg id in request
now in response flow u again use the same algo and create the unique id to get the exact response .. (which ll have the msg id sent as correl id)
and here both the flows are completely independent and not real req res...
the problem we all foreseeing here is the uniqueness of msg id..
is there no possibility of using database for storing the same... or may be in the message body as a field or as a comment in xml or somewhere something...
_________________
Cheers |
|
| Back to top |
|
|
| atommail |
| Posted: Wed Nov 18, 2009 1:47 am Post subject: |
|
|
Novice
Joined: 03 Nov 2009
Posts: 11
Location: Poland
|
| AkankshA wrote: |
so you mean to say that you create a new unique id in request flow and send it as msg id in request
|
No I don't sent msgId in Request flow I have HTTP Request Node I build my order Id and I send my Request
Next In Response in another flow I get my orderId
| AkankshA wrote: |
and here both the flows are completely independent and not real req res...
|
exactly
| AkankshA wrote: |
is there no possibility of using database for storing the same... or may be in the message body as a field or as a comment in xml or somewhere something... |
I don't think so.
So, My idea when I bulid orderIs - use time mor precisely (milisecond) ex CURRENT_TIMESTAMP AS CHARACTER FORMAT 'yyyyMMdd.HHmmssms is bad? |
|
| Back to top |
|
|
| AkankshA |
| Posted: Wed Nov 18, 2009 2:43 am Post subject: |
|
|
Grand Master
Joined: 12 Jan 2006
Posts: 1494
Location: Singapore
|
| atommail wrote: |
So, My idea when I bulid orderIs - use time mor precisely (milisecond) ex CURRENT_TIMESTAMP AS CHARACTER FORMAT 'yyyyMMdd.HHmmssms is bad? |
excuse my ignorance if you think the thread has been dragged too long
well my point is
we are populating the variable orderid with timestamp for its uniqueness
So rather then building the orderid from current_timestamp, why not use the msgId of MQMD as thats reliable enough for uniqueness
i.e. set order id = MQMD.MsgId
_________________
Cheers |
|
| Back to top |
|
|
|
|
