Unsolved
This post is more than 5 years old
7 Posts
0
1274
CICM Steps
Hi Guys,
I have to migrate 2 existing Centera Clusters (6 nodes per cluster) to 2 new ones (4 nodes per cluster).
So I looked at the EMC-Procedure and if I´m right (in "short" steps ->
1) HW Install of the new cube Switches (interconnect with existing cube switches) and 4 Nodes on every site
2) adding the new 4 nodes as a new cube to the cluster (where already one cube exists)
3) move the external role (access) to the new cube
4) start the CICM
My question will be, at point 3 above -> if I have moved the external roles to the new cube, what will happen ?
a) no other roles on the new nodes/cubes are necessary (will the other roles, as storage, automatically migrated to the new nodes?)
b) is the running replication (to the 2nd cluster) affected (after moving the access role) ?
b) is the (read/write) access to the "old" objects on the "old" cube affected (after moving the access role) ?
c) during the CICM, where will be the new objects stored (directly on the new cube) ?
Maybe - if there is another very important point, which I have forgotten,
I would be very thankful for your help and replies!
Thanks & Regards,
Robert
Gotsche_Helmut
48 Posts
0
September 8th, 2014 00:00
Your procedure is correct,
Be aware to move the external role (access, replication, management) only one AFTER the other to have ALWAY at least one nodes with all roles active (regardless from which cube).
When CICM is started, the "available capacity" on the old nodes will be set to Zero, so that all new write will go to the new nodes
Please be aware, that CICM is not a tool to be executed by customer. In case of errors (e.g. node fails) it needs qualified restart.
roboplayer
7 Posts
0
September 8th, 2014 09:00
Hi Mr. Gotsche,
thanks for your fast reply! Our Company is Servicepartner of EMC, last year I have already done a successful C2C Migration; the CICM seems to be easier, but I have still found some obscurities for me (and I want to go sure for a successfull migration ;-)
1) because of the EMC-Procedure (Solve Desktop), maybe it´s not absolutely neccessary to ad a new cube for CICM;
I think for a "partial cube migration" it´s also ok, to add only the 4 new nodes, with storage roles to the existing 6 nodes of cube1; move the access roles one after one from c001n01 & c001n02 to c001n07 & c001n08 ?
look Solve Desktop - Requirements Infos for CICM:
The migration target is all cubes which are not selected as the migration source. In a partial cube migration, it includes those nodes in the source cube which are being kept.
2) the replication should be "disabled" (because of the EMC Procedure) but I think only for the moving action of the access roles ? after I have moved the access roles to the new nodes -> I can activate the replication again (during the CICM proccess) - correct ?
3) in case, if a 2nd new cube (with the new nodes) will be necessary -> a new "pea-file" is not required; the read-access (of the application server for user access) during the CICM will go through the 2nd new cube (access nodes) to the old cube (old nodes) - correct ?
Thanks in advance again, for your support!
Gotsche_Helmut
48 Posts
0
September 9th, 2014 02:00
Ad1
Yes, you can use CICM to "free" some slots in a Centera cluster, assuming you have enough "available capacity" to store the "protected capacity" of the Source nodes.
e.g. perform CICM for c001n01 and c001n02 and after completion install the new 8TB nodes in these slots and then continue with the next CICM
Ad 2
Disable replication (in general: stop Applications) is recommended before moving external roles to be sure, that there is no IO when you unplug the LAN cable.
If you do it ONE after the OTHER, then this ensures, that active applications will always have at least ONE vakid TCP/IP connection from and to Centera Cluster.
Ad 3.
No, you do not need a new PEA file, as this contains only the credentials for the Centera ClusterIDs and the ClusterID is not changed during a capacity upgrade.
It is the advantage of CICM (compared to C2C), that pools and profiles remain valid and no change in the application connection is needed.
roboplayer
7 Posts
0
September 9th, 2014 02:00
Hi Mr. Gotsche,
thanks again for your reply, by the way - I had another chat with EMC-Support and they told me ->
1) a new Cube (for CICM) is necessary, because the 4 new nodes have another disk-size.
2) the replication must be paused/disabled during the complete cicm Procedure ?
(I can´t belive this!) I thought it´s only for moving the external-role action process)
3) I read the cicm-process will get about 5 - 11Mill objects per Day on the new cube and if the Point 2 above is really true,
it will take 2 - 3 weeks (for 180Mill Objects), without active replication ?
Gotsche_Helmut
48 Posts
0
September 9th, 2014 02:00
Before we go into a further theoretical discussion, please contact me directly at helmut.gotsche@emc.com and let me know the serial number or better the recent Health Report of the system you need to tech refresh
ad 1 new cube is only needed if 12TB nodes are used
ad 2 I agree with you;
ad 3 CICM has low priority to minimize the affect for production applications. Do not expect a fast process.