Hi
Sounds like the perfect place for another gridmap. One could use an
interface like ganglia, where by default you get data for the most
recent hour, but there is a drop down menu where you can choose 'hour',
'day', 'week', 'month', 'year'.
One could make a gridmap of all sites, with a temperature-style color
scale representing the %of the period that the site is going to be up
according to the goc db. you could then see the predicted "scheduled
up" state of the grid for the coming hour (ie more or less now), coming
day, coming week, etc.
We might also choose some different, contrasty color for UNscheduled
downtime, this might be helpful.
Given that most of the grid should be up most of the time, the
exceptions should show up rather easily.
JT
Burke, S (Stephen) wrote:
> LHC Computer Grid - Rollout
>> [mailto:[log in to unmask]] On Behalf Of Cyril L'Orphelin
> said:
>> I understand totally your point of view.
>> But if a site which supports atlas is in downtime how can we inform
>> the atlas community without spaming some people ?
>
> As Oxana said, the problem is that most of the time atlas users won't
> care that a service at some random site is down, and there are too many
> mails to read through for the rare important one. If you notice a
> problem you could go back and try to find the mail - but it would
> probably be quicker to look directly on the CIC portal or the GOC DB.
> Maybe the broadcasts could just be limited to the core services? Or as
> someone else suggested you could send a single mail, once a day or even
> once a week, with a summary.
>
> Stephen
|