Sarcouy

for information : This is happening to multiple customers. Even when the vm of pgloader and the vm of mariadb are on the same hypervisor, the network is fine.

Solved, it was a timeout that i didn't notice since Mariadb didn't logged it. Pgloader was running 3 parallel processes so i didn't read properly the logs.

## Reproduced in dev reducing mariadb's write timeout to 5seconds Well i do not understand why this is getting timed out. It seems that pgloader doesn't read it's tcp buffer...

``` Production logs here : 2025-02-28T18:01:40.017000+01:00 NOTICE Starting pgloader, log system is ready. 2025-02-28T18:01:40.043000+01:00 INFO Starting monitor 2025-02-28T18:01:40.050000+01:00 LOG pgloader version "3.6.a94a0a3" 2025-02-28T18:01:40.060000+01:00 INFO Parsed command: 2025-02-28T18:01:40.190000+01:00 DEBUG CONNECTED TO...

# Tcpdump from production MariaDB just acts fine, trying to write until timeout happens, i just can't get why pgloader doesn't read the buffer. ### From pgloader to mariadb ...

I think this step never starts :  request is sent to mariabd but it's not processing into a queue, is it because we lack of workers?

Another run failed : ``` 2025-03-10T01:07:06.272000+01:00 INFO Stopping monitor 2025-03-10T04:00:04.030000+01:00 NOTICE Starting pgloader, log system is ready. 2025-03-10T04:00:04.079000+01:00 INFO Starting monitor 2025-03-10T04:00:04.089000+01:00 LOG pgloader version "3.6.a94a0a3" 2025-03-10T04:00:04.110000+01:00 INFO Parsed command:...

I think the lparallel is not symetric at all. It opens a reader for a table, and should open the writer for this table so reader doesn't timeout, but pgloader...

Here you can see the submit tasks are sequencially called, how can it be not sequencial when scheduler gets tasks to run? ```lisp ;; no Read Concurrency Support detected, start...

@svantevonerichsen6906 you are right he batch concurrency is deprecated by pgloader. I tried again without this option and still get the same result. what i don't understand in pgloader, it's...