FTA agent software rolling upgrade procedure

What is it?

This is the procedure for performing a rolling upgrade to the FTA agent nodes.

It assumes a non-invasive upgrade (i.e. no incompatible changes between transfer-url-copy and transfer-agents).

When to use it?

When you want to upgrade the FTA agents or change their configuration files without causing service downtime.

Procedure

We use in the example the nodes at CERN. The SMS controls are part of the CERN CC environment; substitute appropriately for your site.

The procedure if appropriate for any agent, VO or channel. It should be applied in sequence to all machines running the FTA agent daemons.

Do NOT mark the node for maintenance in SMS

Currently, the agent failover is such tha tyou SHOULD NOT mark the node for maintenance in SMS.

Upgrade the software / config files

Upgrade the software using SPMA.

spma_wrapper.sh

[ edit the YAIM configuration files ]

Rerun the YAIM config tool:

/opt/glite/yaim/scripts/configure_node \
    /afs/cern.ch/project/gd/SC3/SC4/yaim/agents-split-between-fts101-and-fts102/site-info.def FTA

The YAIM script should report the changes (if any) you made to the various agent config files.

Restart agent daemons

Restart all the agent daemons:

service transfer-agents restart

The missed time per agent is limited to around 1 or 2 seconds. All current transfers will continue to run while this procedutre is running. The new software will be active for new transfers started after the agent has been restarted.

Last edit: SteveTraylen on 2007-04-10 - 13:18

Maintainer: GavinMcCance