About POLLING_SCHEDULER

[joyoner]

I have there questions regarding the advanced usage of Miniflux, and I hope the project team and helpful folks can provide some guidance:

1. The difference and collaboration between the two "SIZE" settings:
   WORKER_POOL_SIZE
   Number of background workers to refresh feeds.
   BATCH_SIZE
   Number of feeds to send to the queue for each interval.

My understanding:
BATCH_SIZE is the number of feeds to be polled after each "POLLING_FREQUENCY" interval.
WORKER_POOL_SIZE, I don't quite understand its working principle. What's the difference between having 1 worker and 5 workers?

2. The difference and collaboration between FREQUENCY and SCHEDULER, and their relationship with BATCH_SIZE and WORKER_POOL_SIZE?

POLLING_FREQUENCY
Refresh interval in minutes for feeds.

POLLING_SCHEDULER
The maximum number of feeds polled for a given period is subject to POLLING_FREQUENCY and BATCH_SIZE.
When entry_frequency is selected, the refresh interval for a given feed is equal to the average updating interval of the last week of the feed. The actual number of feeds polled will not exceed the maximum number of feeds that could be polled for a given period.

My understanding:
POLLING_FREQUENCY is a fixed time interval, such as polling every 10 minutes.
POLLING_SCHEDULER (entry_frequency) adjusts the current polling frequency based on the feed's update frequency from the previous week. However, I'm not sure how this dynamic frequency differs from the POLLING_FREQUENCY mentioned above. Could you provide an example to illustrate?

3. About ENTRY_FREQUENCY's MAX_INTERVAL and MIN_INTERVAL
   MAX_INTERVAL: Even if the feed's update interval from the previous week is greater than MAX_INTERVAL (meaning it doesn't update frequently), it will still be executed at the MAX_INTERVAL frequency.
   MIN_INTERVAL: Even if the feed's update interval from the previous week is less than MIN_INTERVAL (meaning it updates frequently), it will still be executed at the MIN_INTERVAL frequency.

In summary, I have carefully read the manual several times and also conducted practical configuration tests. I summarized the following rules based on the following configuration, but I still cannot explain the three questions above:
POLLING_FREQUENCY=10min
ENTRY_FREQUENCY_MAX_INTERVAL=20min
WORKER_POOL_SIZE=10
BATCH_SIZE=200
BTW, my actual number of feeds is 140+.

On the Web-Feeds page, there are three types of Last check fields:
Last check: 6 minutes ago
Last check: 16 minutes ago
Last check: 26 minutes ago

According to the BATCH_SIZE=200 configuration, in theory, all feeds should be updated every 10 minutes, so there should only be one type of Last check. However, as shown above, there are three types.

[shizunge]

WORKER_POOL_SIZE decides how many feeds it polls simultaneously.

Considering the following case. You have 5 feed. POLLING_FREQUENCY=1 and BATCH_SIZE=5 and WORKER_POOL_SIZE =1

The websites are really slow and it will take 1 minute to finish the polling.

Since WORKER_POOL_SIZE is 1, it will take 5 minutes to poll all feeds. And your effective polling interval is 5 minutes.

If you change WORKER_POOL_SIZE to 5, the polling will be done in 1 minute.

If you change WORKER_POOL_SIZE to 2, the polling will be done in 3 minute.

POLLING_FREQUENCY plays two roles. We would better to separate them. One is how often we add feed to the queue. One is SCHEDULER_ROUND_ROBIN_INTERVAL_MIN (I believe this is due to a bug). See #2201 (comment)

POLLING_SCHEDULER does not adjust the POLLING_FREQUENCY, but it adds more criteria whether a feed should be added to the queue. As a result, Miniflux won't poll some feeds too often.

Did you use a new database every time you change the parameters?
Previous setting may affect the result of next experiment, if you did not start from a fresh database.