Considering to switch away from KDEPIM/Akonadi

[Martin Steigerwald]

Hi.

I did not write in a long time about own issues with KDEPIM/Akonadi and 
instead tried to help others along here and there, but now I feel quite 
frustrated…

also due to own resistance to actually report a bug, I deal with a 
regular crash of KMail¹ – regular as in several times a day.

However, that is not the main issue.

The main issue are two performance related issues (despite MariaDB 
tuning and all that):

1) During mail retrieval KMail sometimes just stops displaying new mails 
coming in. It GUI is responsive, but when I click on another folder or 
so… I just get the waiting message. These stalls, that do not happen all 
of the time, take up to a minute or even longer.

2) Filtering of new mail into folders takes very long. Very long means 
up to 5 minutes or move of sustained activity of both Akonadi and 
MariaDB for filtering 1000-2000 new mails. That is with me even making 
sure to remove old mails from Akonadi's maildir entirely. Most folders 
do not have more than 30000-40000 mails, some considerable less. A few 
have 80000-100000 mails and I have the idea to remove mails out of them 
as well.

3) Then I had one of the LRCONFLICTs displayed in KMail's status line 
without even able to expand it to see what is is about. I solved it by 
clearing Akonadi cache of incoming folder.

At the same time due to introduction of Office 365 at work² I partly use 
Evolution with EWS and I found the following:

1) Evolution did not crash *once* so far.

2) Evolution just works. For me so far, that is.

3) Evolution does not get stuck.

3) So far I do not see much of a performance issue. There are some short 
stalls at time, but nothing even almost as unbearable as with KMail and 
Akonadi.

Granted the amount of mails in my work-around is considerably lower. And 
while KMail does not crash several times a day with my work setup, it 
still crashed once in a while. Also Akonadi gets stuck with IMAP or EWS 
connection or whatever else.


Now, I know how much I defended KDEPIM / Akonadi, but… I also see the 
amount of effort you, Daniel, and others put into fixing things up and 
even changing Akonadi in some fundamental ways in order to make it work… 
I truly appreciate that work…

but I am also tired.

Simply put: I am fed up with it.

Yes, I see the recent improvements in Akonadiconsole to help others 
report meaningful bugs and all that…

but I do not like to be a database admin or a server administrator just 
to read my mail and respond to them or write a bug report.

KMail is still *just* a mail program.

Akonadi does not have issues since yesterday. I'd say it had issues till 
the beginning. I might have been lucky at times where it mostly worked 
for me, but also I moved through a ton of issues and pain with KDEPIM 
and Akonadi.

I am sorry to write but I have found myself seriously considering to 
switch away from KDEPIM, as much as I like KMail.

The only other idea I had is to just switch from MariaDB to PostgreSQL. 
It would mean a PostgreSQL upgrade once in a while, but it if works 
better, I'd probably do that. Especially after I somewhere skimmed that 
you, Daniel, wrote about a kind of locking issue with MariaDB/MySQL. And 
also since you, Daniel, as the only one I am currently aware of tackling 
deep into Akonadi, AFAIK use PostgreSQL as well.

I am reluctant as akonadictl fsck reported a lot of issues the last time 
I used it, including about items without RID. I may like to export those 
into files before attempting a switch as Akonadi still cannot reply 
those. And still I also do not get how a maildir resource could ever 
fail to write out the mail into the maildir as long as the filesystem is 
healthy.  I am also reluctant due to having to reconfigure about 100 or 
more filter rules.

Or I might give it another change till the rewrite of the search 
infrastructure as one of the last steps to improve Akonadi's 
architecture.

Or at least wait till 18.12 or 19.04 packages are available and test 
again.


But as much as it saddens me I am quite close to just being done with 
it… and take the huge step to migrate away from KDEPIM / Akonadi.

How long has Akonadi been in development? Way more than 5 year, even a 
decade probably? It ought to be stable.

Yet, just by monitoring KDEPIM bugs mailing list I know it is far from 
being stable.

Yes, I know I have argued in favor for attempting to fix up Akonadi… and 
maybe its really close now with only one of the major steps in the major 
architectural overhaul missing – new search infrastructure.

And yes, I'd like to be convinced to hold out a little bit longer, but 
on the other hand I did so again, and again, and again…

and it still does not work as stable and robust as I'd like it to. I 
still put up with issues that a software that is being developed for 
more than 5 years – correct me if my memory is wrong – simply should not 
have anymore.

So maybe, just maybe it is important to start admitting that Akonadi is 
still not in the state it is supposed to be… and consider all options…


I end it now, its already long enough. I am preparing for a recreational 
time out. I will most likely be offline or mostly offline from Friday to 
Monday.

I bet it might spur quite a discussion or so… all I ask of you is to 
stay friendly in tone. I truly see that KDEPIM developers are working 
hard, just AFAIK there are only a few who dare to dig into Akonadi. Dan 
and well David recently.

I hope that this "rant" while honest is still friendly in tone. I still 
remember walking around angrily in this room, venting "such a piece of 
crap software" at quite some times. In hind sight I do not mean it that 
harshly. I see the good ideas behind Akonadi as well, but the 18.08 
implementation of it, feels like a major burden to me. 17.12 has been 
better, as it did not have the stalls during receiving mails and KMail 
did not crash several times a day. But it still used a lot of resources 
and mail filtering was still slow. However it was a lot more bearable 
than 18.08 appears to me.


[1] Bug 404052 - Crash during/after filtering inbox, probably related to 
Qt WebEngine integration
https://bugs.kde.org/404052

[2]  I do not agree with that move, but that is how it is currently…

Thanks,
-- 
Martin