This post is more than 5 years old
6 Posts
1
46110
August 3rd, 2009 03:00
In Navisphere host connecting to EMC CX300 are showing unmanaged
Hi All,
I have EMC CX300 storage connected to two servers PE2950. Both the server are accessing the storage LUNS and installed the powerpath 4.6 and Navisphere Agent also configured on the servers. All the redundant four paths are showing in both servers but still in Navisphere the host are showing with "U" mark and says Host Agent are not reachable.
How this can be removed, please help to solve the issue.
Copper .
No Events found!
BaseerM
6 Posts
1
August 3rd, 2009 08:00
Thanks Dev, can you forward me the link to upgrade the powerpath from 4.6 to 5.2.
Also, previously I have experience that once I upgraded the powerpath from 4.5 to 5.0 and the server doesnt recognise/read the LUNS so I roll back and works fiine again. Is it very important to upgrade or can I manage with this only and if it is important to upgrade then please mention if any care or measures to be taken before upgrade.
As you said these two servers were in cluster before, but now there is no cluster at all but still the LUNS are shared between them and accessible to both. Also we are facing a problem in using that LUNS. Whenever we are trying to install Oracle database and giving the path M:\ drive (storage) it is giving error and installation stops in the middle. After trying a lot we could able to install the database on M:\ drive but now we cannot access any file from this drive. What ever files we try to open it says "directory or file corrupted and unreadable". This is the main issue for which I started to look the storage and found unmanaged host. Please help.
BaseerM
Dev Mgr
4 Operator
•
9.3K Posts
1
August 3rd, 2009 12:00
To upgrade from Powerpath 4.5 or older to 4.6 or newer, the best way is to back up your license key (copy and paste it into a text doc, write it down, or whatever), then uninstall Powerpath 4.5 or older, reboot and then install 5.2SP1 (and provide the key when prompted).
Up until Powerpath 4.5.x Powerpath used EMC's MPIO driver. As of 4.6.x it switched to Microsoft's MPIO framework (only on Windows servers obviously). This change is best done 'clean' instead of via an upgrade.
When 2 servers share a LUN whose filesystem isn't cluster aware (e.g. VMFS, GFS, OFS2, but not NTFS, EXT3, EXT4, etc), when one server writes to the disk, the other doesn't know that space is now used, so it can start writing to the same space and overwrite the first server's data causing data corruption and all sorts of problems.
This is a SAN limitation (not just CX300, but any SAN (iSCSI, FC or SAS)). MSCS would only allow 1 server (node) at a time access to the filesystem, which is how it keeps the filesystem 'safe' from corruption.
VMFS (VMware ESX's native filesystem to storage Virtual Machines (VMs)) allows only the server that own the VM write access to the folder where the VM is stored, preventing the other servers from writing to the data in that folder. This is how VMFS can safely be shared among multiple ESX servers.
I'm not familiar with the internal workings of GFS and/or OFS2 to explain how those cluster-aware filesystems can safely allow multiple servers access to the disk.
tommo666
1 Rookie
•
1.2K Posts
0
August 3rd, 2009 03:00
Check the agent.config file in notepad
There should be 2 entries;
system@ ip of spa and sytem@ip of spb
You can also create a txt file called agentID.txt in c:/ consisting of 2 lines
1st line 2nd line once created either restat the agent service or reboot the server.
Dev Mgr
4 Operator
•
9.3K Posts
0
August 3rd, 2009 07:00
Also, the "U" (unmanaged) is completely harmless.
Just two things; you say "Both the server are accessing the storage LUNS", and you reference Powerpath 4.6.
I'd recommend first to upgrade to a newer version of Powerpath (e.g. 5.2 SP1), and keep in mind that 2 servers can only share LUNs if the servers are clustered (e.g. Microsoft Cluster services, Red Hat Cluster, VMware ESX cluster, etc).
BaseerM
6 Posts
0
August 3rd, 2009 08:00
Thanks Quote for the reply... For the two entries you said "system@ ip of spa and sytem@ip of spb", I already added them at the time of installing agent on the servers.
Also I tried to restart the agent services but of no use.
Copper
tommo666
1 Rookie
•
1.2K Posts
0
August 3rd, 2009 13:00
If the servers are multi homed, ie more than one ip assigned to the nics then the agent may try to communicate with the CX through one of the nics not on the network like a heartbeat link for clustering, which results in the 'U'. AgentId.txt specifies which interface to communicate through via the hostname and the ip of the public interface. It works for Linux and MS servers.
BaseerM
6 Posts
0
August 4th, 2009 01:00
Thanks Dev, I will clean and install the fresh upgrade version, please forward me the link to get upgrade of powerpath.
Also, FYI our environment is MS WIndows 2003 (Ent) with v24 of Navisphere, so what is your opinion shall I disconnect the LUN from one of the server and I can solve this issue OR disconnect from both and add freshly to one of the server to solve, please need your great advice.
Baseer.
BaseerM
6 Posts
0
August 4th, 2009 01:00
Thanks Dev, OK I will clean install the fresh upgraded version, please forward me the link.
BaseerM
6 Posts
0
August 4th, 2009 01:00
Thanks Tom, the heart beat link is no more connected now, the servers are with single NIC and connected through the LAN and it is pingable to the storage proccessors. FYI, the servers are MS Windows 2003 (Ent) and Navishpere is v24.
Dev Mgr
4 Operator
•
9.3K Posts
0
August 4th, 2009 07:00
To get Powerpath, log into Powerlink (powerlink.emc.com) and under the software downloads you can find Powerpath.