I'm having some difficulty with dnews at the moment in that it keeps deleting items from groups which it shouldn't.
This has happened to me once before. The symptom is that, all of a sudden, items just vanish. Every single new item is deleted immediately, and it takes a rebuild_index for it to start storing new items again.
Once the rebuild_index is complete, functionality returns. However, the groups in question are then missing several thousand earlier posts.
On closer examination of dnews.log, rebuild_index finishes, but it is during resort_index that the problems arise.
29 08:58:32 :info: cmdq: Processing next background command (resort_index) 29 08:58:33 :info: Index Sort chunk size is 300000 items 29 08:58:33 :warn: mem: dbmmem malloced RESORT 29 08:58:33 :warn: Processing cancelled items, this could take some time 29 08:58:33 :info: db: cancel upto 0 29 08:58:33 :info: Cancel processing 0 buffered cancels, upto 0 29 08:58:33 :info: db_clear_cache(cancel_background) 29 08:58:33 :info: Group clear cache 29 08:58:33 :info: Group clear cache - done 29 08:58:33 :warn: Checking index for consistency (nightly check) 29 08:58:34 :info: group: Group Write S:\dnews\work/active.new 29 08:58:34 :info: group: Group write done 29 08:58:35 :info: Index check completed 29 08:58:35 :info: db_clear_cache(db_reindex) 29 08:58:35 :info: Group clear cache 29 08:58:35 :info: Group clear cache - done 29 08:58:35 :info: Sorting dbi.idx into 5 stacks 29 08:58:35 :info: db_reindex_new_scan 0 29 08:58:35 :info: Pruning large group demon.local:29349 29 08:58:36 :info: Pruning large group alt.alien.vampire.flonk.flonk.flonk:475386 29 08:58:36 :info: Pruning large group demon.local:2387 29 08:58:36 :info: Pruning large group uk.media.tv.misc:16 29 08:58:36 :info: Pruning large group demon.local:3686 29 08:58:36 :info: Pruning large group uk.media.tv.misc:4268 29 08:58:36 :info: Pruning large group demon.local:5340 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10695 29 08:58:36 :info: Pruning large group demon.local:5361 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10929 29 08:58:36 :info: Pruning large group demon.local:5388 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10938 29 08:58:36 :info: Pruning large group demon.local:5389 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10940 29 08:58:36 :info: Pruning large group demon.local:5419 29 08:58:36 :info: Pruning large group uk.media.tv.misc:11085 29 08:58:36 :info: Pruning large group demon.local:5432
The logs then fill up with thousands more lines stating that items from groups have been "pruned". Yet, I do not want it to "prune large groups" as I have my expire.conf deliberately set on very large retentions. There is no problem with space...
I haven't come across that message before, it is in recently added code (ChrisP). A brief look suggests it is related to the max_items setting in dnews.conf. This specifies the maximum number of articles DNews can store per group. Exceeding this value could cause problems, so the code was most likely added to prevent the value being exceeded.
Set max_items in dnews.conf, I personally have set this to 1000000 on my own server. You must restart DNews after making this change as it alters internal buffer sizes.
> I'm having some difficulty with dnews at the moment in that it keeps > deleting items from groups which it shouldn't.
> This has happened to me once before. The symptom is that, all of a sudden, > items just vanish. Every single new item is deleted immediately, and it > takes a rebuild_index for it to start storing new items again.
> Once the rebuild_index is complete, functionality returns. > However, the groups in question are then missing several > thousand earlier posts.
> On closer examination of dnews.log, rebuild_index finishes, but > it is during resort_index that the problems arise.
> 29 08:58:32 :info: cmdq: Processing next background command (resort_index) > 29 08:58:33 :info: Index Sort chunk size is 300000 items > 29 08:58:33 :warn: mem: dbmmem malloced RESORT > 29 08:58:33 :warn: Processing cancelled items, this could take some time > 29 08:58:33 :info: db: cancel upto 0 > 29 08:58:33 :info: Cancel processing 0 buffered cancels, upto 0 > 29 08:58:33 :info: db_clear_cache(cancel_background) > 29 08:58:33 :info: Group clear cache > 29 08:58:33 :info: Group clear cache - done > 29 08:58:33 :warn: Checking index for consistency (nightly check) > 29 08:58:34 :info: group: Group Write S:\dnews\work/active.new > 29 08:58:34 :info: group: Group write done > 29 08:58:35 :info: Index check completed > 29 08:58:35 :info: db_clear_cache(db_reindex) > 29 08:58:35 :info: Group clear cache > 29 08:58:35 :info: Group clear cache - done > 29 08:58:35 :info: Sorting dbi.idx into 5 stacks > 29 08:58:35 :info: db_reindex_new_scan 0 > 29 08:58:35 :info: Pruning large group demon.local:29349 > 29 08:58:36 :info: Pruning large group alt.alien.vampire.flonk.flonk.flonk:475386 > 29 08:58:36 :info: Pruning large group demon.local:2387 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:16 > 29 08:58:36 :info: Pruning large group demon.local:3686 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:4268 > 29 08:58:36 :info: Pruning large group demon.local:5340 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10695 > 29 08:58:36 :info: Pruning large group demon.local:5361 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10929 > 29 08:58:36 :info: Pruning large group demon.local:5388 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10938 > 29 08:58:36 :info: Pruning large group demon.local:5389 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:10940 > 29 08:58:36 :info: Pruning large group demon.local:5419 > 29 08:58:36 :info: Pruning large group uk.media.tv.misc:11085 > 29 08:58:36 :info: Pruning large group demon.local:5432
> The logs then fill up with thousands more lines stating that > items from groups have been "pruned". Yet, I do not want it > to "prune large groups" as I have my expire.conf deliberately > set on very large retentions. There is no problem with space...
>I haven't come across that message before, it is in recently added code >(ChrisP). A brief look suggests it is related to the max_items setting >in dnews.conf. This specifies the maximum number of articles DNews can >store per group. Exceeding this value could cause problems, so the code >was most likely added to prevent the value being exceeded.
Ahh. I've just checked the DNews manual based on what you've told me and this does appear to be the problem. The manual says that the default setting is 90,000, and I believe that the groups in question had reached 90,000, which would have triggered the "prune".
>Set max_items in dnews.conf, I personally have set this to 1000000 on my >own server. You must restart DNews after making this change as it alters >internal buffer sizes.
Given that you have your max_items set much higher than 90,000, I think it is safe to assume that I shouldn't have problems if I use a similar value to you!