Re: Akonadi server won't start
Martin Steigerwald <[email protected]>
| Newsgroups | gmane.comp.kde.users.pim |
|---|---|
| Message-ID | <2984648.p44xP0n9gS@ananda> |
Hi David. David Jarvie - 19.08.21, 01:35:05 CEST: > On Tuesday, 17 August 2021 20:26:00 BST Ingo Klöcker wrote: > > WARNING: Switching the database will make Akonadi forget more or > > less > > everything. Make a backup before proceeding. If you have lots of > > filters configured, then export them before the switch. If you have > > mail stored in local folders, i.e. if you are not using IMAP > > exclusively, then you might also want to export those folders. > > Akonadi may not have written all local mail to local files so that > > some of those mails may only be stored in the database. > > I've already encountered the issue which you describe, i.e. that after > moving mails from IMAP into local folders, no file is created in the > directory for that folder. I've tried copying instead of moving, but > that doesn't work either (unlike in older KDE PIM software). > > Is there any way of forcing mails in the Akonadi database to be > written to local folder disk directories? It really isn't > satisfactory for them to only exist in the database. I am not aware of any such way. It has been topic here, in bug reports and I think also on kde-pim mailing list before. On top of that for items that Akonadi failed to write to the remote storage, be it IMAP or a local maildir folder, remote is anything that is not the Akonadi cache, Akonadi will not attempt to write them at a later time. You see those with "akonadictl fsck". See those Item "1744643" in collection "205" has no RID. Item "1744697" in collection "205" has no RID. Item "1744696" in collection "205" has no RID. Item "1744764" in collection "205" has no RID. Item "1744770" in collection "205" has no RID. Item "1744836" in collection "205" has no RID. Item "1744866" in collection "205" has no RID. Item "1764092" in collection "205" has no RID. Item "2123315" in collection "205" has no RID. Item "2123303" in collection "205" has no RID. messages. Usually after a short while I had quite some of these in every Akonadi setups I have. Even those that except some very low volume IMAP accounts (a few mails a month if at all) have only local maildir storage. So far I have never got why Akonadi could fail to write out a mail file to local storage unless there would be an issue with the filesystem like it being 100% full. Maybe it also happens if you stop or restart Akonadi at the wrong time, but I never really understood this. Maybe in case you stop it during a long running background operation. Still Akonadi is only able to do one operation at a time. So an folder indexing operation can block out KMail for a longer time. Often enough I stopped KMail and then Akonadi with "akonadictl stop" cause I wanted to be able to continue my mail work. Thing is Akonadi is conceptionally a cache, however it extends that concept also to being a write cache. That means, before writing out an item it goes through the Akonadi cache. That means there is the chance that there is something in the cache that is not stored elsewhere so far. Idea is to be able to deal with IMAP servers that are unreachable at times. However for that Akonadi would need to retry storing the changes onto the IMAP server at a later time, yet this is not implemented. Evolution on the other hand does not close the mail composer window until it made sure that the mail is stored on IMAP or EWS account. But it also displays a message when you try to close it during background operations. KMail does not need to wait for that as Akonadi operates in the background. So Akonadi theoretically is more advanced, but in my point of view it is unfinished. And maybe a bit too complex and thus error prone. And I still claim it does a lot of needless work, likely re-scanning an IMAP folder after every 5 mails you delete. It would be good at some point to actually tackle those shortcomings. Daniel Vrátil did some work at this level, but he kind of burned out it seems before finishing his "make indexing great again" work that would at least avoid moving every mail through the cache for indexing. I do not understand enough of Akonadi to actually address these shortcomings. And that probably part of the challenge here: Finding someone who is able to, willing to and has enough time to do the ground work that is needed to finally stabilize Akonadi to a point where it is rock solid and snappy at all times or replace it with something different. Plus there should not be a need for "akonadictl fsck". No other mail application I know needs this. Evolution does not, Thunderbird does not, mutt does not, notmuchmail does not. Or if it can't be avoided, it should be able to at least attempt to write those items without RID to remote storage again, but currently it just tells the other about the inconsistency between cache and remote storage without doing anything about it. That is at least how much I understood. I still stick with KMail cause it is a really good application. However I wondered to switch to Evolution often enough. I use it at work cause Akonadi with EWS is just not working reliable enough for me. Best, -- Martin