kotori
kotori copied to clipboard
daq-tools
Fix "Connection pool is full"
A simple configuration option while creating the InfluxDB client should be enough. https://influxdb-python.readthedocs.io/en/latest/api-documentation.html#influxdbclient
pool_size (int) – urllib3 connection pool size, defaults to 10.
Added here
Dear Adrián,
thanks for writing in and for using Kotori. May I ask you how you encountered this problem? While I know there might still be issues with the connection pooling, I am eager to get some repro or more detailed information about this, especially how it manifests for others.
With kind regards, Andreas.
Just a pet project of an arduino sending data each 12.5ms
El sáb., 21 mar. 2020 a las 18:17, Andreas Motl ([email protected]) escribió:
Dear Adrián,
thanks for writing in and for using Kotori. May I ask you how you encountered this problem?
With kind regards, Andreas.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/daq-tools/kotori/issues/28#issuecomment-602074746, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAYWPGAFUAEIMBNESDV75Z3RITZDZANCNFSM4LQ6PW4Q .
Thanks for the details. So, you are just sending a high throughput of datapoints to the MQTT bus which quickly triggers this condition and you are observing this through log messages on STDERR?
I will see what I can do against that. Do you believe that raising the pool size will solve the problem? Have you been able to solve it that way? And if so, which pool size did you use?
I am just thinking about what will happen if the pool is exhausted again. Shall we think about (additionally) introducing queues in order to buffer the writing?
Yes, I increased the pool to 50 and it worked without warnings. I think just giving the option to users could be enough.
El sáb., 21 mar. 2020 a las 20:16, Andreas Motl ([email protected]) escribió:
Thanks for the details. So, you are just sending a high throughput of datapoints to the MQTT bus which quickly triggers this condition and you are observing this through log messages on STDERR?
I will see what I can do against that. Do you believe that raising the pool size will solve the problem? Have you been able to solve it that way? And if so, which pool size did you use?
I am just thinking about what will happen if the pool is exhausted again. Shall we think about introducing queues?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/daq-tools/kotori/issues/28#issuecomment-602089916, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAYWPGGLGCDQMRA46DNRRATRIUHA3ANCNFSM4LQ6PW4Q .
All right, thanks! I will add pool_size as an option to the InfluxDB configuration then.
Dear Adrián,
your suggestion has been implemented by b3cb7344 and will be part of the next release. Thanks again!
With kind regards, Andreas.