Don't stop batch runs when one of the hosts is not available

[thekorn]

I you do a fix node:all and one of the nodes inbetween is not available the run will stop immediatly with an error - which is extremly annoying.

to prevent this from happening we can use env.skip_bad_hosts = True The only remaining question is if it is safe to set it to True in all cases, of if this needs an commandline option, with False (the current behaviour) as default?

[thekorn]

@tobami, any idea on that one?

[tobami]

When using the concurrency option failures are ignored, so we have a start there. I would not make it default in a first version, as that changes assumptions and may break something for some users.

I would call the variable skip_failures. The implementation I guess would be to catch the errors in the node runner

[thekorn]

the problem here is that - if we only change this environment variable to True - we won't really implement a skip_failures feature, it is just a skip_on_connection_error but not a skip on chef config errors.

Do you still think the naming makes sense, even with this tiny change, or do you want a full implementation of this feature?

[tobami]

OK, let's begin with skip_on_connection_error