Dear All
Please note this from and provide feedback to Philippa or myself.
Alessandra's comments in response yesterday are below for reference to
the only feedback given so far from UKI.
Many thanks,
Jeremy
For inclusion in the ops manual:
"SE downtime declaration
- If a site has no SEs available, it should be declared as down
- If only some of the SEs are not available, the downtime should be
declared for those SEs and not for the site
- When a site puts some of its SEs in to downtime but still has some
remaining they
should configure on their WNs that the environment variable
VO_<VO>_DEFAULT_SE should point to one of the SEs still in operation"
Alessandra's response:
"I didn't see your request for comment before.
> - If a site has no SEs available, it should be declared as down
Many application do not need a SE. It is a waste of resources putting in
downtime a whole site if the SE is down.
> - When a site puts some of its SEs in to downtime but still has some
> remaining they
> should configure on their WNs that the environment variable
> VO_<VO>_DEFAULT_SE should point to one of the SEs still in operation
I consider temporary changes on any node but in particular on WNs, a
last resort in case of emergency that shouldn't be inserted in
operational guidelines.
There should be another way to allow VOs interested in using local SE to
distinguish sites. For example it could be included in the FCR. VOs
could easily extract the information from the GOC DB about local SE
availability, otherwise as a sys admin why do I waste my time inserting
the information there. Or they should have some redundance mechanism in
their jobs. I know some users do have them, and they are not even
production users."
|