Simon Marechal
Simon Marechal
@semharel Not sure I understand the question, but if it is about the configuration file format, I expect an object where keys are header names, and values are header values.
This is also an old bug : the current field for header action makes no sense as a single string. The configuration should be as described in the message from...
I had to put back configuration auto reload to keep compatibility with the current mode of operation, and I saw some problems related to that, as with the async code...
Also this is a POC, logs are not stored in elasticsearch for now, so the first test will always fail.
I'll try to fix the logging issue tomorrow so that people can see something on elasticsearch, but it will not have the same format as the previous log, for now.
And there is no grasshopper support!!!! This explains the remaining failures.
You should run the tests against configuration that use Redis (rate limiting / flow control), as this is that is supposed to be a showstopper in terms of scaling capability.
Scratch can't really be made thread-local, as we are moving to an async model where this will not work well, but I'll take a look right now at the performance...
That was well spotted, in my tests scratch allocation accounts for 25% of total run time (when there are no Redis calls, in the present of Redis calls this becomes...
It seems that a workable design would be to have a "matcher" task that would hold a single scratch and would run all the matching. This would not be a...