This is an issue reported by Marian Zvada. I created a ticket for it since we should at the very least discuss it:
during setup of new factory we've hit the following case:
- once we put all entries "down" (without any record in
glideinWMS.downtimes), we are not able to bring entry back "up", it
fails. I guess it might be related to non-existent record about specific
entry down rather that record that "all" entries are down.
I was able reproduce this behaviour, few step how one can do it are
listed below [*].
Don't know whether it should be classified as bug or something which
might be documented as "feature" if someone installs factory from
scratch and tries to migrate tens of entries from the other factory.
If one do downtime particularly per entry, it is indeed possible bring
it back "up" w/o hassle.
#1 Updated by John Weigand about 8 years ago
Understand that I am not at all familiar with how downtime is implemented. My question is "why is does this just affect newly 'newly' installed factories?". It would seem like this would/could affect any factory to which new entry points are being added in mass (and maybe just one perhaps?)
#3 Updated by Igor Sfiligoi about 8 years ago
Here is my take on this:
What Marian is describing is not a bug, but a "feature".
We want to have a way to put the whole factory down, and bring it back up, while preserving the various downtime states of the various entries.
Please make sure this functionality is preserved.
However, Marian pointed out an interesting new use case;
when installing a new factory, one may indeed want to proceed cautiously and start with all entries down.
And then bring up one at a time.
However, I don't think this should be implemented in the "manageDowntimes" tool.
It should be instead a feature of create_glidein/reconfig_glidein; i.e. if so desired, new entries are created in downtime state already.
#6 Updated by Douglas Strain almost 8 years ago
Here is my proposal:
This functionality would CHANGE:
./factory_startup down -entry all [or -entry entries]
would now be the same as doing "./factory_startup down -entry X" for each entry X. This would destroy any combination of entries up/down you have. All entries would be set as in downtime. Conversely,
./factory_startup up -entry all [or -entry entries]
Would be the same as doing "./factory_startup up -entry X" for each entry X. All entries would then be up and out of downtime regardless of what they were previously.
However, I propose that the following stays the same as before:
./factory_startup down -entry factory
This would set the factory as being down, and leave all the entries alone. This could be used to preserve entry-specific downtimes while setting the whole factory as down and then later as up.
Does this solve the problems? If so, I will code this later on this week.