vcs patch

[tanislavm]

Hi,

I like to verify things.

If i have 2 nodes vcs with applications running on them,and i wish to patch or upgrade VCS,then it is enough to stop the vcs on both nodes and apply patches or upgrade.Right?Then restart VCS.

This is possible because the OS and VCS files on disks are different from application files on disks.Right?Normally OS and VCS files are on local disks and application is on SAN.Right?So if i patch or upgrade the VCS files,this will not overlap any of disk application files nor OS files on disk.

Then initial installed OS files on disk are different from VCS files on disk.Right?

The same if i like to add a new node to VCS,it is enough to stop vcs on nodes within cluster,then have in sync(edit) the llttab,llthosts,gabtab and main.cf on all the nodes,so i will next start vcs on all the nodes.right?

On Solaris if i wish to patch or upgrade VCS,i will use in this way-i create an liveupgrade BE,i will patch and upgrade it.

To conclude.The OS layer is below of VCS layer and this one is below of application layer.I take those things like different entities,and the OS files are not overlapped by VCS files on disk.

Any application runs at user mode.Right?

Thanks so much.

[Gaurav_S · Accepted solution]

Here are my responses

If i have 2 nodes vcs with applications running on them,and i wish to patch or upgrade VCS,then it is enough to stop the vcs on both nodes and apply patches or upgrade.Right?Then restart VCS.

>>>>  Yes, you would need to stop VCS & apply patches. Whether you need to shutdown apps or not, this depends on patch or upgrade you are making. If you doing a major upgrade from an older version, you might need to reboot which may ask for outage. however if its a minor upgrade, you can shutdown only vcs using "hastop -all -force" which will leave applications running. You can then patch/upgrade VCS & restart VCS. You may want to keep applications running on other node while you patch first node so that impact can be minimized. Once first node is done, offline everything on second node & online on first node. Then patch your second node.

This is possible because the OS and VCS files on disks are different from application files on disks.Right?Normally OS and VCS files are on local disks and application is on SAN.Right?So if i patch or upgrade the VCS files,this will not overlap any of disk application files nor OS files on disk.

>>> Yes, application binaries are different from VCS binaries. These files are different. If you upgrade VCS, VCS binaries & files are modified. Some of startup scripts for VCS may be modified as well at OS layer but this will not harm OS functionality. In a typical scenario, application binaries are kept on local storage but application data is in shared storage (SAN or NAS) but this may vary from customer to customer.  VCS files will be typically on local disks.

Then initial installed OS files on disk are different from VCS files on disk.Right?

>> Yes, OS files are different from VCS files.

The same if i like to add a new node to VCS,it is enough to stop vcs on nodes within cluster,then have in sync(edit) the llttab,llthosts,gabtab and main.cf on all the nodes,so i will next start vcs on all the nodes.right?

>>> Yes, you can prepare your new node completely (install VCS, configure VCS by taking correct llthosts, llttab, gabtab, main.cf , fencing etc), once done, you can stop cluster & seed gab with additional node. You will need to modify gabtab file on other nodes as well to startup with new nodes if cluster reboots.

On Solaris if i wish to patch or upgrade VCS,i will use in this way-i create an liveupgrade BE,i will patch and upgrade it.

To conclude.The OS layer is below of VCS layer and this one is below of application layer.I take those things like different entities,and the OS files are not overlapped by VCS files on disk.

>> Partially correct, top most layer is application, below that is VCS, below that will be OS

Any application runs at user mode.Right?

>> Yes, typically specific users are configured to run application, many of applications are capable to run with root user only.

G

[tanislavm]

Hi Gaurav,

Thanks and i like some clarification.

You state:"You may want to keep applications running on other node while you patch first node so that impact can be minimized. Once first node is done, offline everything on second node & online on first node. Then patch your second node."

Fine,but if the vcs id down and applications are up,how i move the application from a node to other node.manually stop the application on running node and manually start it on the other node?

So when i add a new node to vcs,after all the configuration files are in sync on all nodes,then i stop vcs and leave the applications running,then i restart the vcs on all nodes.Right? 

I state:"On Solaris if i wish to patch or upgrade VCS,i will use in this way-i create an liveupgrade BE,i will patch and upgrade it."

Is this the best way to minimize the downtime?Because normally OS and VCS are on locally disk.

[Gaurav_S]

Fine,but if the vcs id down and applications are up,how i move the application from a node to other node.manually stop the application on running node and manually start it on the other node?

>> two ways, either before shiutting down VCS, move applications to other node or if VCS is shutdown, start apps manually on other node.

So when i add a new node to vcs,after all the configuration files are in sync on all nodes,then i stop vcs and leave the applications running,then i restart the vcs on all nodes.Right? 

>> correct, by vcs, meaning you will need to shutdown entire stack of vcs, vxfen, gab & LLT

I state:"On Solaris if i wish to patch or upgrade VCS,i will use in this way-i create an liveupgrade BE,i will patch and upgrade it."

Is this the best way to minimize the downtime?Because normally OS and VCS are on locally disk

>> yes, live upgrade is pretty good method to minimize downtime.

G