Well, the two ways are completely decoupled … I hope this is clear.
The syslog way is not available in 5.26/00xxx (for a time this is documented … ): it was introduced after 5.26/00 . But it has been ported to 5.26/00-proof, so, if you want, you can give a try from there.
The fixes/changes to the SQL way will also be ported to 5.26/00-proof .
Not yet. I could not work on this much yet, unfortunately … Also, the changes are probably going to be non-backward compatible for Monalisa (which is the only implementation really used currently; by ALICE), so I am trying to understand what is the why to minimize the impact.
In addition, I will be on holiday next 10 days, so it is likely that nothing will happen before the week of August 23rd.
could you please post it here, if you manage to fix it in the 5.26-proof patches? I was just interested in the sql monitoring part, Monalisa looks like too “big” for our local proof farm, I just needed a simple accounting and from the sql database I can easily extract plots and tables with accounting and monitoring info. So it would be great If those fixes could go to the 5.26-proof branch, for we might not be ready to move to 5.27 so soon.
well, in the end I continued debugging
I suspected that the crash I was getting was somehow related to the reserved keyword “group” issue.
I thought when TSQLMonitoringWriter::SendParameters creates the sql sentence “INSERT into blabla”, it should frame the group colum name between quotes as well ``.
That was one step forward, but still I had to remove single quotes from the column name, i.e. (user) instead of (‘user’).
With those fixes, the SQL monitoring works fine (so far) with my mysql database.
I guess all these are the issues with the sql monitoring you already detected, but just in case I leave here the patch I applied.