Hello,
Our database has grown - a lot - and because of this success we are hitting some major performance issues.
I saw in the application logs we are now hitting regularly some errors from MySQL on lock wait timeouts.
Of course performance is always based on many different aspects, but regarding the application / configuration side, would you have any suggestions to get around this ?
I am thinking about a few things , at least short term:
1- increase the lock wait timeout value for MySql database. Perhaps giving it more time to complete transactions will be enough.
2- the timeouts I've seen are mainly on insert or delete into 'kpihistory'. So I'm thinking of removing the Key Indicators as a temporary workaround.
3- I'm also hoping we can reorganize some legacy projects. Some projects have been going on for years, and have a sub-project tree hierarchy which is quite huge. It seems to be mainly users on this type of project who hit the most performance issues.
Any other suggestions would be welcome ! 🙂
I also believe I saw that the upcoming major release has some performance improvements - would that perhaps help large legacy projects ?
Regards
1) not the best option I think
2) I recorded ticket #5122 on that point. I think it could be interesting to un-synchronize these updates. Just store request for recalculate, then recalculate when no new uptate for a given period...
3) possibily consolidatio nto top level could be un-synchronized also. More dangerous, but could be effichient.