Fix send failed due to message too long

We periodically get the following error in the logs:

"message":"In DriveHandlerProxy::addLogParams(): Socket send failed."
"bufferLength":250207

Note that this already contains the bufferLength as extra debug info. It seems that the watchDog parameters are not flushed frequently enough, causing them to pile up. They should be flushed more frequently (but not too frequently) to remove this error.

added cta: Tape server priorityhigh typebug labels

assigned to @nbugel

created branch 1025-fix-send-failed-due-to-message-too-long to address this issue

mentioned in merge request !755 (merged)

mentioned in issue #1028 (closed)

For now I have added an extra flush at the end of the read task, but we end up getting a lot of the stillOpenFileForThread messages, so that is suspicious. Trying to debug this further

assigned to @afonso

Due to the (relatively) slow flushing of the parameters to the parent process, we ended up flushing inside the same message requests for both adding and deleting the same parameters ("stillOpenFileForThread[...]" params):

Examples:

• https://gitlab.cern.ch/cta/CTA/-/blob/main/tapeserver/castor/tape/tapeserver/daemon/DiskReadTask.cpp#L78
• https://gitlab.cern.ch/cta/CTA/-/blob/main/tapeserver/castor/tape/tapeserver/daemon/DiskReadTask.cpp#L169

!755 (merged) prevents this, by not flushing params that were both added and deleted.

As a result, we don't expect more than one "stillOpenFileForThread[...]" param being sent at a single time, which should keep the message under it's max length.

closed with merge request !755 (merged)

mentioned in commit 6d90c943

mentioned in merge request !844 (merged)

mentioned in commit 767a78ef