Author |
Message
|
vsethi007 |
Posted: Tue May 11, 2010 5:37 am Post subject: Cluster Problem in case of servers shutdown |
|
|
Novice
Joined: 17 Apr 2008 Posts: 19
|
We have a majot shutdown planned for all the servers we have.
All MQ servers will be down and we have lots of Qmgrs which are in clusters.
All servers will bought up after 2days...will servers shutdown can have a severe effect on clustering.
I have little idea on Clustering but what I know is once cluster is up it'll try to set channel between Full repositories but is connection on first attempt cannot be made then there can be severe problem and we need to take Qmgrs out of cluster then recreate and bring them back into cluster..is that true...
What precautions or proactive steps we should do to avoid any complications |
|
Back to top |
|
 |
bruce2359 |
Posted: Tue May 11, 2010 5:42 am Post subject: |
|
|
 Poobah
Joined: 05 Jan 2008 Posts: 9469 Location: US: west coast, almost. Otherwise, enroute.
|
A cluster is a collection of qmgrs that have objects defined as part of one or more clusters.
If you are shutting down ALL of your servers, then there will be no problems at all
If you are shutting down SOME of your servers, AND some are part of the cluster, then you need to ensure that the Full Repository qmgrs remain alive, AND that all Partial Repository qmgrs have CLUSSDR to the Full Repositories that are to stay alive during the outage. _________________ 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 |
|
 |
Vitor |
Posted: Tue May 11, 2010 6:38 am Post subject: Re: Cluster Problem in case of servers shutdown |
|
|
 Grand High Poobah
Joined: 11 Nov 2005 Posts: 26093 Location: Texas, USA
|
vsethi007 wrote: |
I have little idea on Clustering but what I know is once cluster is up it'll try to set channel between Full repositories but is connection on first attempt cannot be made then there can be severe problem and we need to take Qmgrs out of cluster then recreate and bring them back into cluster..is that true... |
Not even slightly true.
The "cluster" won't come up because there's no such thing. As individual queue managers come back up they'll attempt to re-establish the channels already defined. This includes the manual ones pointing to the Full Repositories. If these are down, the channel will go into retry until the Full Repository queue manager comes up.
The biggest problem you're likely to face is that no changes (new queues, new queue managers, etc) can be made to the cluster until at least one of the Full Repositories is active. So you should plan to have one or both of them started as early as possible (or ideally not both down at the same time). The rest of the cluster will behave normally & indeed any workload distribution will take account of the fact some of the queue managers are unavailable.
So it should all just work. _________________ Honesty is the best policy.
Insanity is the best defence. |
|
Back to top |
|
 |
vsethi007 |
Posted: Tue May 11, 2010 7:15 am Post subject: |
|
|
Novice
Joined: 17 Apr 2008 Posts: 19
|
hanks for your replies..
The only thing here is since all servers will go down so we don't have possibility of keeping one Full repository up....everything will be down for 2 days...
Do we need to follow any sequence for startup of servers...I mean should servers having Full repository should start first followed by partial ones or they can start in any order...
By biggest worry is what can be possible disastrous situation for which I should plan for if you can through light on that.....as nothing is going to change its just a normal shutdown for 2 days for all servers....then all will be back up. |
|
Back to top |
|
 |
exerk |
Posted: Tue May 11, 2010 7:19 am Post subject: |
|
|
 Jedi Council
Joined: 02 Nov 2006 Posts: 6339
|
Start both Full Repositories first. The Partial Repositories can then be started, and in any order. _________________ 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 |
|
 |
bruce2359 |
Posted: Tue May 11, 2010 7:27 am Post subject: |
|
|
 Poobah
Joined: 05 Jan 2008 Posts: 9469 Location: US: west coast, almost. Otherwise, enroute.
|
But even starting the FR's first is not a requirement. PRs channels will go into retry until the FRs come to life. _________________ 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 |
|
 |
Vitor |
Posted: Tue May 11, 2010 7:43 am Post subject: |
|
|
 Grand High Poobah
Joined: 11 Nov 2005 Posts: 26093 Location: Texas, USA
|
vsethi007 wrote: |
Do we need to follow any sequence for startup of servers...I mean should servers having Full repository should start first followed by partial ones or they can start in any order... |
As I said above, you should start one or both of the Full Repositories first.
vsethi007 wrote: |
By biggest worry is what can be possible disastrous situation for which I should plan for if you can through light on that |
Well disasters should have their own plan and you should make sure that's up to date.
vsethi007 wrote: |
.....as nothing is going to change its just a normal shutdown for 2 days for all servers....then all will be back up. |
The only likely problem you're going to have is if neither of the Full Repositories back up for 90 days and some of the unusued auto-defined channels will fail. But if these servers are not back after 90 days you'd be well served using that DR plan I mentioned above.....  _________________ Honesty is the best policy.
Insanity is the best defence. |
|
Back to top |
|
 |
vsethi007 |
Posted: Wed May 12, 2010 1:32 pm Post subject: |
|
|
Novice
Joined: 17 Apr 2008 Posts: 19
|
Thanks for all your replies....
I have one more doubt on it..
Since I have couple of servers and each servers hosts a queumanager on it and all qmgrs of every servers are in cluster...
Now I have 5 Qmgrs which host full repositoy and 2 of them are on Windows systems which is using DHCP, we have used hostname in CONName parameter of cluster sender/receiver channels...
But when restart will occur DHCP will assign new IP to servers and even proving hostname is not a relaible approach...will it have a -ve impact when a servers will bought up? |
|
Back to top |
|
 |
Vitor |
Posted: Wed May 12, 2010 1:37 pm Post subject: |
|
|
 Grand High Poobah
Joined: 11 Nov 2005 Posts: 26093 Location: Texas, USA
|
vsethi007 wrote: |
Now I have 5 Qmgrs which host full repositoy |
But not more than 2 in a given cluster I trust
vsethi007 wrote: |
even proving hostname is not a relaible approach |
Why not? Using hostnames rather than actual IP addresses is basic networking and should be completely reliable. If it's not, it's a network issue not an WMQ one. _________________ Honesty is the best policy.
Insanity is the best defence. |
|
Back to top |
|
 |
shashivarungupta |
Posted: Tue Jun 15, 2010 4:16 am Post subject: |
|
|
 Grand Master
Joined: 24 Feb 2009 Posts: 1343 Location: Floating in space on a round rock.
|
vsethi007 wrote: |
..
Now I have 5 Qmgrs which host full repositoy and 2 of them are on Windows systems which is using DHCP, we have used hostname in CONName parameter of cluster sender/receiver channels...
But when restart will occur DHCP will assign new IP to servers and even proving hostname is not a relaible approach...will it have a -ve impact when a servers will bought up? |
As per Dynamic Host Control Protocol the IPs would be (re)assigned to the servers , but I believe most of the applications used to Connect to the qmgrs using Hostname(port) not the IP(port).
If you got any confusion what DHCP would do then please consult your network admin guys.
As far as Server Down Process is concerned, it would not disturb the cluster and its objects until you make any changes that would put you in some difficult situation. As Vitor exerk and bruce said that all will work fine if you would follow given instructions.
Why don't you do some TEST on your local machine by creating the cluster(s) and then bringing down/up the server(s).
 _________________ *Life will beat you down, you need to decide to fight back or leave it. |
|
Back to top |
|
 |
aditya.aggarwal |
Posted: Wed Jun 16, 2010 1:30 pm Post subject: |
|
|
 Master
Joined: 13 Jan 2009 Posts: 252
|
vsethi007 wrote
Quote: |
All servers will bought up after 2days...will servers shutdown can have a severe effect on clustering. |
is it in production ? DO You Have DR servers in place?
Do you know the affect on Business if the Mq is down for two days?
Hope you have right approvals for this activity!!! |
|
Back to top |
|
 |
|