-
|
It seems like that if we have multiple delta fuseki server, the individual fuseki will not fetch the newest updates from the patch log server until someone make a request to it. For example, if I have a load balancer and 3 delta fuseki servers behind it, when I make a update request to one of the servers, the other two servers will not pick up the update unless there is a sparql request route to themselves. When this happens, the querying will be very slow since the fuseki need to catch up the newest update. |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 4 replies
-
|
Sending a harmless query will work. Something like |
Beta Was this translation helpful? Give feedback.
-
|
Does patch log server expose any API that allow user to check the current version of patch? This way we only need to send a small query when there is an update instead of constantly making calls to the server. |
Beta Was this translation helpful? Give feedback.
-
|
Check the current version - yes - but that isn't enough to determine whether an update is going to happen. That needs to know the state of the Fuseki server - it may be several versions behind. Pinging it with If you do want to avoid the regular ping, have the update request touch each server after the update returns. |
Beta Was this translation helpful? Give feedback.
-
|
Great! That makes sense, thank you |
Beta Was this translation helpful? Give feedback.
-
|
BTW, out of curiosity, what is the endpoint that patch log server expose that allow user to check the current version? |
Beta Was this translation helpful? Give feedback.
-
|
Interface All operations except patch push/get (called |
Beta Was this translation helpful? Give feedback.
Check the current version - yes - but that isn't enough to determine whether an update is going to happen. That needs to know the state of the Fuseki server - it may be several versions behind. Pinging it with
ASK{}does that calculation and is cheap. If, say, once every 5 seconds, it should be negligible.If you do want to avoid the regular ping, have the update request touch each server after the update returns.