You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When the name-server starts, it parses the pvDirectory.txt file and associates all the PV names from each file listed with that IOC. The code linked below is the part of the connection handler that runs when the IOC's heartbeat PV successfully connects:
There's an assumption in that code which is wrong. An IOC that was down when the name-server was restarted might write a different set of PV names to the file when it next boots, adding or removing names. The name-server marks each IOC as up when the heartbeat PV connects, but the code there assumes that it doesn't need to reload the PV names on the first connection (which could be hours or days later).
A fix which handles the initial connection exactly like the reconnect case would resolve this, but restarting a running name-server would then parse the file from each active IOC twice. That's undesirable for large sites like APS (750 IOCs, 1.75m PVs), so I want to stop it reading all the files at startup, allowing that parsing to happen as each IOC's heartbeat connects.
The text was updated successfully, but these errors were encountered:
When the name-server starts, it parses the pvDirectory.txt file and associates all the PV names from each file listed with that IOC. The code linked below is the part of the connection handler that runs when the IOC's heartbeat PV successfully connects:
ca-nameserver/src/main.cc
Line 725 in 709e1fe
There's an assumption in that code which is wrong. An IOC that was down when the name-server was restarted might write a different set of PV names to the file when it next boots, adding or removing names. The name-server marks each IOC as up when the heartbeat PV connects, but the code there assumes that it doesn't need to reload the PV names on the first connection (which could be hours or days later).
A fix which handles the initial connection exactly like the reconnect case would resolve this, but restarting a running name-server would then parse the file from each active IOC twice. That's undesirable for large sites like APS (750 IOCs, 1.75m PVs), so I want to stop it reading all the files at startup, allowing that parsing to happen as each IOC's heartbeat connects.
The text was updated successfully, but these errors were encountered: