Yes, it was really helpful. I’ll provide you an updated Olympia version today.
Hi guys,
Just to keep you updated. More than 36 hours straight without locks in Olympia. So far so good.
We are getting them again. 
Any suggestions? It is less than before but they will probably increase with more traffic.
A good thing is that they are not frequent. So, nothing radical is needed to alleviate it yet e.g. 3 today, but 2 of them could be related to the same collision. Need to check if there was an olympia restart or crashed that may cause a sudden increase.
Hi guys,
Just as an update. Now that volume of deadlocks is increasing and we are getting 5 to 7 locks per day. Sql that is causing the lock:
It seems that the cleanup operation conflicts with other data update operations. Guess I’ll have to do something similar with the event subscriptions fix and pause execution of any data update operations during the cleanup phase. This needs some investigation ofc.
Thank you Anton.
Important notes, today in our dev environment everything crashed.
All services that depend on event subscriptions start failing. Restarting Olympia will not restore things. Services will start sending messages for OnConnected/OnDisconnected one after the other one.
Checking Olympia Db I can see 34 sessions. Only 12 of them were actually still valid, the rest were at least days old and Olympia didnt clean them up.
EventData had 646219 rows.
EventSubscriptions 456 rows.
EventUser 456 rows.
EventUserData 4146439 rows
All clearly wrong. So the clean up process on Olympia is not working and its taking everything down
I keep an additional db, where all my sessions, subscriptions and additional session data is being tracked so I dont depend on Olympia and they all report no maximum currently of 28.
Our dev environment works 24/7 and its used by a lot of services. I can only imagine whats going on in production. Problem is, I depend on Olympia right now and it is not doing that well.
Is there a way to exposed the different elements, like GetAllSessions, AddSession, RemoveSession, GetEventSinks so it can be replaced by calls to a different service so I can attempt to plug a service that will do similar actions to Olympia?
Existing implementations exposed ways to access the data sources, stored procedures, etc, but they expect those components to be inside of the same service. I will like just to have the calls so I can plug calls to another service out of process.
Could you send me that database, given it is dev-only and does not contain sensitive data?
You can import Olympia’s API using Service Builder from address supertcp://localhost:8011/bin
Sorry about that Anton, too late. Olympia DB tables were cleaned up in order to be able to start things up in our environment. I typed this while I am dealing with the protests from other places.
Until now I’m able to partially start things up. I can do post mortem analysis. I’m now going to get the row counts in production, I’m afraid of what I’m going to find. I’ll post back.
It that’s a cleanup code issue then rows will add up again in dev environment and in a day or two there will be a new test dataset.
Ok, now i need help.
Thats production environment. More than one million records in several of them.
I need a way to get rid of the excess without taking everything down.
What will be a valid sql script that i can execute on Olympia to clean this up.
Also, the number of locks is increasing which I’ll guess is related somehow with this.
thank you.
