replication server

[bof]

Hello all,

I have 2 L1000 servers that I would like to sinchronize the data and os once a day. What would be the best way of doing that.

Note that the primary server must not stop working.

For the moment the solution that is in place goes as follow (I do not fond it very nice)

The main server has 3 disks in mirror. Once a day a routine detaches one of the diks and start copying the data to the ohter machine (takes 2-3 hours). Once it is finished the disks re-attaches it self (takes for ever to re synch).

What is a good cheap way of doing that ?

[TwoProc]

You didn't say whether or not this is a database server, or if it is, what type.

If this isn't a database server, then the easiest way is just to use "rsync". It's not included in HPUX, but can be obtained from the porting archives (search google for "HPUX porting archive"), and run through an ssh pipe easily enough.

If this is a database type of server, then the answer lives in the "it depends" world.

[bof]

not it is not a DB.
I will lokk for the rsync

[bof]

Let say I use resync

I would like to copy all disks OS disk included.

What would be the best way to figure out if the config of one server will work on the other. the 2 servers do not have the same harware version, processor, etc and other small diffrences.

[TwoProc]

Are you sending compiled executables? If so, the best way is by performing a good round of testing to see if the application(s) work as expected... unfortunately, no shortcuts there.

But, that's if the procs are LEAST in the same family (that is PA-8800s, PA-8700s,etc). If they are not (for example the source machine is an HPUX box and destination is an Intel Linux box) of the same processor family, then this won't work. You'd have to rsync only the data areas, and leave the compiled code behind. Also, the data areas may or may not be compatible if going from 64 to 32 bit, big endian to little endian architecture... in other words, there are as many answers and non answers as there are combinations of source machines and destination machines,operating systems,etc.

The answer to this question: "it depends."

[bof]

That's what I expected.

As I dig a bit more into this I am now puzzeled to see that the script does a copy of 20 G every night since it copies all the data every night ...

[TwoProc]

Well, that's why rsync is so nice, if this is a collection of files, some (or many) of which don't usually change, then rsync is much more efficient than copying the whole thing over every day. Rysnc will only copy over the changed files in a directory tree, based on file contents and/or date/time stamp. It can optionally delete files as well. It gives a nice report of which files it is sending or deleting to make the directories sync up.