Rebalancing replication resource

[olga.me...@gmail.com]

Hello,

I'm playing with replication resource in iRODS 4.0.3. I wonder if I always should run 'iadmin modresc myReplResc rebalance'  command to start replication process or there is an automatic way for it in iRODS (rules ?!)?

Are there any best practices (guidelines) for a resource tree organization? If I create one replication resource (with two children each on a separate resource server) in the beginning what is the minimum change I will have to do to scale it out in the future?

Thank you for your help,
Olga

[Terrell Russell]

Hi Olga,

By default, the replication resource will keep all its children in sync - a file put into the replication resource will create a replica in every child resource.

You should only need to run the rebalance subcommand when you do some 'surgery' to the resource composition (aka, add a new child resource which doesn't have any existing data objects).  The minimum change to 'scale' out your deployment would be to add enough storage - the policy you implement is independent from the storage you add/manage under iRODS.

How you organize your tree will be driven by your use case.  If you want to always have multiple replicas, then the replication resource is for you.  If you want to have certain resources be read-only, you could use a weighted passthru to change the 'voting' coming/going to each child.  If you want to keep things evenly distributed, you could use a random or round-robin resource with multiple children.

Terrell

--
--
"iRODS: the Integrated Rule-Oriented Data-management System; A community driven, open source, data grid software solution" https://www.irods.org
 
iROD-Chat: http://groups.google.com/group/iROD-Chat

---
You received this message because you are subscribed to the Google Groups "iRODS-Chat" group.
To unsubscribe from this group and stop receiving emails from it, send an email to irod-chat+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[olga.me...@gmail.com]

Hi Terrell,

Thank you for the answer. It was my mistake actually as I did use an empty file to test the replication :-))) and as result it was on one machine and was not on the other.

As about the tree structure.. I just don't know about the data size/usage yet. I thought I could start from something very simple/generic and then probably restructure it in the future. Your answer added a bit of confidence to my decisions.

Thank you,
Olga

[Terrell Russell]

Ah, an empty file was an optimization in the older code - sorry for the edge case.

In the upcoming 4.1.0 and later, all files, including empty files, will trigger the replication code and be resident on all the children of a replication coordinating resource.

Terrell

--