by alexon » Sat Nov 20, 2021 3:11 pm
Thank you Quade!
I will definitely try to follow your recommendations.
Unfortunately, I cannot fully follow your recommendations on limiting "Display Age", since my specifics of work require periodic return to messages created several years ago.
I also have to say: I deeply respect You for such a responsible approach to your work. It's rare now.
Added:
I noticed the following: the set of groups to which I am subscribed has not changed for many years.
The size of the spool_v6 folder located on the NVMe SSD drive is approximately 11 GB.
The size of folders inside it varies from 1 MB to 900 MB.
In version 682, loading headers from the database was much slower than in RC5.
In both versions 6.82 and RC5, at the end of the process of creating a list of headers, the load on the CPU increases dramatically (the processor liquid cooling system howls).
But in the case of 6.82, the list of headers is formed successfully, but in the case of RC5, the program collapses without even leaving entries in the logs.
Maybe some software does not have time to process such a large data stream coming from NVMe SSD? It can be an antivirus or NB or a Windows indexer or something else.(maybe I'm wrong)
I think so, because even if there are no new messages in the binary group, but it has, for example, 600,000 messages, then RC5 crashes in the process of creating a list of headers,
but version 682, although slower, correctly creates a list of headers.
Another example: a binary group has 300,000 messages.
Process of creating the list of headers, in both RC5 and 6.82 works perfectly.
Added:
Your recommendations helped.
Set the NB display age to 3000 days. Everything works. Then I set the value to 3500 days - the NB crashes.
In version 6.82, I can read messages from 2008 and earlier.
Last edited by
alexon on Sat Nov 20, 2021 5:08 pm, edited 1 time in total.