EqualLogic firmware upgrade

In the FW Package there is a Readme with some important notes about STOPPING  existing Replication Schedules and so on.

Yes, you will upload the FW Package through the GUI (cmd also possible) and it will be loaded into the standby controller. If this step is finished it will be wait for the restart command. On this step the failover will occur and the standby controller becomes activ. It depends on your Switch configuration how long it will take to get the ports into admin mode. With portfast enables it will takes less 10seconds.

During this period your Servers cant reach their volumes and they need some proper configuration to survive this. But if Hit KIT, VMware Tools for VMs, MeM is installed on the Host all settings mostlikly already set. In an VMware environment you have to set the Initiator Login Timeout from 5 up to 60 seconds.

You can "ping" mgmt and storage port IP if you like to see whats going on. For years now we performing an Online FW Upgrade and never have a problem in an VMware Environment (arround 20 arrays). All Volumes stay online.

If you have only one member... you will lost the connection to the Groupmanager. It will be reconnect automaticly so dont get nervous.

Normally these kind of test is part of the deployment and can be initiate by pressing the "restart" button on the maintanence tab. So you can be sure that you have a propper cabling and Switch/Host configuration.

Have the serial cable around for .... well you know ;)

Things which can be wrong:

- No space left because of some crash dumps or diags. You may have to clean up first by log into through SSH.

- Once i got a msg about problem for updating the flash cards (its running in the background and can take some time). I got in contact with EQL support to sort this out. Updating from the cmdline returns more detailed informations.

Regards,

Joerg

We use hyper-v vms and it is necessary to suspend them before upgrade?

Did I understand it right that after we start upgrade process:

1. passive controller will be updated

2. passive controller will be rebooted

3. storage LUNs fail over from active to passive

4. active controller becomes passive and will be updated and reboot

5. all this will happen during same upgrade process

Could we jump from 5.2.5 straight to 6.0.5?

thanks

You are correct on the upgrade process.  Yes, you can go directly from 5.2.x to 6.0.5, no intermediary steps needed.

With the firmware download is a PDF "OS Considerations" guide.  That has info on setting the disk timeout value for Windows and Linux servers.   All your VMs and the Hyper-V hosts need this registry value set.  Not just for firmware upgrades, but to also handle a controller failover due to a HW failure.  If the entry is not set, you will have to reboot your VMs and hosts to make the change effective.

Doing upgrades during a slow IO period or maintenance window is helpful.   My personal prefernece is to use the serial port for upgrades.  I use a terminal server so I can monitor both serial ports, watching the process carefully.  Additionally, if there was a problem, the serial port could be only access method.  having it setup and working means a quicker resolution.   Lastly, if you have multiple members in a group, especially in the same pool, stage the restarts.  Restart one, wait for CLI message (or in GUI) that the member is back online.  Wait a minute or so, then restart the next one.

Regards,

As I don't have account now with me, but :

"With the firmware download is a PDF "OS Considerations" guide.  That has info on setting the disk timeout value for Windows and Linux servers."

Is this regarding just 'TimeOutValue' in "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Disk". As on Windows 2012 I see it is default 60s value (3c). Is this correct value recommended by Dell? Any other considerations?

Thanks

That's the biggest one.  Yes.

Note that if you install the hitkit, it also sets the disk timeout value to 60 if it wasn't already on that.

This disk timeout does not apply to VM's running regular disks inside the VM right?, just talking VM's using HIT/ME volumes right (like a SQL or Exchange)?  Just might want to clarify for newbies

.

Side note to folks getting ready to load 6.0.5, we had to re-install SAN HQ 2.6 after going from 5.2.6 to 6.0.5, not sure why, kept failing after loading java jar file around 70% when logging into member from SAN HQ gui.  They need to scrap Java yesterday and go with HTML5 or other new language.

We do like the snapshot space borrowing feature.

It is best to also set the disk timeout value to 60 seconds on virtual machines. If the virtual machines have the hitkit installed on them (e.g. for guest-attached-iSCSI), it has already been taken care off.

It helps the virtual machines survive a raid controller failover.

I would say it's CRITICAL to set the disk timeout value to 60 on any server, VM or physical.   The lower timer inside the VM might expire and cause the OS drive to crash.  Even though the ESX host does not return an error to the VM.  There are two times when the controller will failover.   When you want it to and when you don't expect it to. Being able to handle the controller failover over for Firmware upgrades also mean you can deal with the unexpected CM failover events.

Hello,

The problem is not the version you are jumping to.  As the message indicates you can do the upgrade via the GUI with only one controller installed.   GUI upgrade will use failover of the controller to process the upgrade.   With only one controller the entire unit is rebooted.  

So upgrade to 7.1.8 either via SSH session to that member, or preferably via the serial port.

In this case you don't have a passive controller, so the reboot will take longer.   So yes, I would shutdown your VMs and ESXi servers.  

Regards,

Don