--- Log opened lun ene 17 00:00:37 2022 00:53 < fikabot_> 💬 > <@hank:myportal.social> And latest MySQL doesn't have the NO_AUTO_CREATE_USER setting either...posted about that in the issue 00:54 < fikabot_> 💬 00:54 < fikabot_> 💬 Although Oracle MySQL and MaridDB share a common ancestor, they now go in different directions. All development on Friendica is done with MariaDB, so it is likely that we don't automatically catch up with compatibility changes that had done by Oracle, but we depend on the reports of the users who are using MySQL. 00:58 < fikabot_> 💬 But also we try to reduce the complexity of our SQL statements in a constant preparation to eventually support databases line PostgreSQL. 08:13 < fikabot_> 💬 how are the chances that Friendica will ever see PostgreSQL support? 08:19 < fikabot_> 💬 I wouldn't hold my breath, but I wouldn't give up hope, either 09:27 < fikabot_> 💬 I'm still waiting for Sqlite support. Even Matrix Synapse has it 09:30 < fikabot_> 💬 I'm a much bigger fan of Postgres so like to hear that, hypothetically 09:30 < fikabot_> 💬 I nstalled MySQL because that's what the install docs mentioned :( MySQL 5.6+ or an equivalent alternative for MySQL (MariaDB, Percona Server etc.) 09:31 < fikabot_> 💬 should I try dumping the DB and reinstalling with it 09:31 < fikabot_> 💬 > <@bkil:grin.hu> But this also seems like a complete guide: https://www.howtoforge.com/how-to-install-friendica-social-network-platform-on-ubuntu-20-04/ 09:31 < fikabot_> 💬 09:31 < fikabot_> 💬 This guide showed you to install the mariadb package, though. 09:32 < fikabot_> 💬 So I would advocate of rephrasing the whole install.md guide at every place, that mariadb must be installed and mysql needs to be put in bracket instead if that is what the developers are testing on (similarly to why PHP 7 is being installed). 09:32 < fikabot_> 💬 That guide is a bit better in some ways 09:32 < fikabot_> 💬 but the create your own site stuff didn't work when the server itself is on a subdomain rather than managing the whole domain 09:32 < fikabot_> 💬 > <@hank:myportal.social> does that config change need to be added to the install instruction troubleshooting? 09:33 < fikabot_> 💬 09:33 < fikabot_> 💬 I think that would be nice even if it was just the error message, the config line that fixed your issue and a link to the issue. 09:33 < fikabot_> 💬 I ended up punting back to putting it directly into /var/www/html to get it working 09:34 < fikabot_> 💬 and using the pre-make the local.config.php and running the CLI rather than the website method...not sure why that didn't work in this case but it didn't 09:34 < fikabot_> 💬 that may have been because Apache was still using PHP 8 when I Thought it was using 7 09:35 < fikabot_> 💬 Maybe. Although I think Friendica kind of works on PHP 8 as well.. 09:35 < fikabot_> 💬 > <@bkil:grin.hu> I think that would be nice even if it was just the error message, the config line that fixed your issue and a link to the issue. 09:35 < fikabot_> 💬 09:36 < fikabot_> 💬 Would that be a "if you are using MySQL rather than Maria DB" thing now though? Swapping the language on the DBs in the requirements as you suggested would be primary 09:36 < fikabot_> 💬 > <@bkil:grin.hu> Maybe. Although I think Friendica kind of works on PHP 8 as well.. 09:36 < fikabot_> 💬 09:36 < fikabot_> 💬 Yeah once it was up and running I had no problem 09:40 < fikabot_> 💬 Overnight it built up quite the worker log, I think from the importing...with *tons* of errors about MySQL going away etc...I was trying to run it on a 1GB memory system. Perhaps that was insufficient? 09:40 < fikabot_> 💬 I bring that up to ask if minimum system requirements should be in the install along with software requirements 09:42 < fikabot_> 💬 I think if it runs out of memory and gets shot down with 128MB-256MB RAM, it should be reported. Your logs would also have evidence of this. 09:43 < fikabot_> 💬 MySQl logs don't say why it was crashing 09:43 < fikabot_> 💬 it just has the startup messages 09:43 < fikabot_> 💬 Do you see OOM messages in syslog? 09:43 < fikabot_> 💬 oh damn I forgot DigitalOcean doesn't change the Ubuntu default of not having swap 09:43 < fikabot_> 💬 Do you see OOM messages in syslog/dmesg? 09:43 < fikabot_> 💬 Do you see OOM messages in syslog/dmesg/logread? 09:45 < fikabot_> 💬 Yeah it was killing MySQL for using too much memory 09:46 < fikabot_> 💬 https://github.com/friendica/friendica/issues/3351 09:46 < fikabot_> 💬 Oh, okay then. 09:46 < fikabot_> 💬 I think you could also reduce the buffer & sizes of MySQL to help this. 09:47 < fikabot_> 💬 How many cores does the server have? Perhaps it might benefit if you reduced the count of the allowed number of parallel workers/queries. 09:47 < fikabot_> 💬 I originally tried 1 vCPU 1 GB RAM 09:47 < fikabot_> 💬 https://tech.labelleassiette.com/how-to-reduce-the-memory-usage-of-mysql-61ea7d1a9bd?gi=9db59ea5217f 09:47 < fikabot_> 💬 the workers are mostly idle 09:47 < fikabot_> 💬 even when there are a lot of them 09:48 < fikabot_> 💬 But yeah, if you have the option to add swap, you might want to add a little. 09:48 < fikabot_> 💬 I usually do but forgot 09:48 < fikabot_> 💬 I *hate* that they don't configure swap out of the gate now 09:48 < fikabot_> 💬 Sure, the workers might be idle because they are waiting for either the network or the database. 09:48 < fikabot_> 💬 yup 09:49 < fikabot_> 💬 I guess my point was it didn't have some weird behavior because the system was too busy 09:49 < fikabot_> 💬 > <@bkil:grin.hu> https://tech.labelleassiette.com/how-to-reduce-the-memory-usage-of-mysql-61ea7d1a9bd?gi=9db59ea5217f 09:49 < fikabot_> 💬 09:49 < fikabot_> 💬 https://dev.mysql.com/doc/refman/8.0/en/memory-use.html 09:49 < fikabot_> 💬 > <@bkil:grin.hu> https://dev.mysql.com/doc/refman/8.0/en/memory-use.html 09:49 < fikabot_> 💬 09:49 < fikabot_> 💬 http://www.tocker.ca/2014/03/10/configuring-mysql-to-use-minimal-memory.html 09:49 < fikabot_> 💬 I'm not intending to reconfigure MySQL but thanks for the link 09:49 < fikabot_> 💬 https://bobcares.com/blog/fix-mysql-high-memory-usage/ 09:49 < fikabot_> 💬 links 09:50 < fikabot_> 💬 you can stop now :) 09:50 < fikabot_> 💬 It was in my worker queue 😉 09:50 < fikabot_> 💬 I imagine this is an artifact of trying to do my import 09:51 < fikabot_> 💬 Well, actually, if you enabled MySQL query debugging, it might be useful to see whether the issue was more about caching, a large amount of queries in parallel or an unoptimized query that could be fixed in Friendica. 09:51 < fikabot_> 💬 I could try that 09:51 < fikabot_> 💬 But if you can solve this with swap, sure that would be less expensive, just saying. 09:52 < fikabot_> 💬 Believe me, this path of the code is not tested well (importing huge, real life profile timelines) I think at least this should also be mentioned in both install.md and move-account.md. 09:52 < fikabot_> 💬 Also lack of statusing for the importing user 09:53 < fikabot_> 💬 I can tell it's generating tons of work sstill 09:53 < fikabot_> 💬 (queue built up to 1000 over night with the db issues) but from the "user" perspective it looks completely done 09:53 < fikabot_> 💬 still no friends in my list though :( 09:54 < fikabot_> 💬 What is the intended artifact from a connected user's perspective when it is done? 09:54 < fikabot_> 💬 I think you could open a github issue to ask that (along with also giving this hint in move-account.md that only the queue size shows the ongoing progress status) 09:54 < fikabot_> 💬 Like if bill@friendica1.com is friends with maria@friendica2.com and bill moves to newfriendica.com will suddenly maria see her friend bill at "bill@newfriendica.com"? 09:54 < fikabot_> 💬 > <@hank:myportal.social> What is the intended artifact from a connected user's perspective when it is done? 09:54 < fikabot_> 💬 09:55 < fikabot_> 💬 I don't know, but I hope I can finally found out from you after you succeed 😄 09:55 < fikabot_> 💬 But in testing what did you all find? 09:55 < fikabot_> 💬 I never tested this! 😊 09:55 < fikabot_> 💬 Yeah but to others in the room who did 09:56 < fikabot_> 💬 I dabbled with it in D* back when I was developing on it but I don't remember a thing about it lol 09:56 < fikabot_> 💬 never did it on a real account between real servers 10:00 < fikabot_> 💬 > <@bkil:grin.hu> I never tested this! 😊 10:00 < fikabot_> 💬 10:01 < fikabot_> 💬 PS is it safe to assume that if the import doesn't go well and I wanted to do it again it'd be as simple as deleting the user that I did it through and trying again? 10:01 < fikabot_> 💬 Where said delete would also delete all of the post content, but probably not federated user data, associated with it? 10:09 < fikabot_> 💬 Well, I'm not sure. I would wait for others to chime in. But if it was me, it would not harm the original account until you press `Nuke me` on that one. 10:10 < fikabot_> 💬 Does deleting a user delete purge their content though? 10:11 < fikabot_> 💬 IIRC the D* account migration said you only got one shot...but that may have been assuming you didn't have an administrator who could delete the user in a way to let you start again 10:11 < fikabot_> 💬 And it's been since 2018 so my memory may be hazy 10:17 < fikabot_> 💬 On the daemon thing...has someone attempted setting this up as a systemd thing or something that ensures there is always one process running? 10:21 < fikabot_> 💬 On migraton...seeing worker tasks like "UpdateContact 2836" is probably a good thing in terms of notifying contacts I've moved. What that means practically is TBD :) 14:35 < fikabot_> 💬 I saw an AddContact command as well 14:35 < fikabot_> 💬 still no contacts yet though... 14:36 < fikabot_> 💬 I saw an AddContact command in the queue just now as well 14:42 < fikabot_> 💬 I thought you tried to replicate the process in a sandbox environment just yesterday. Wasn't this the scenario you tested? Setting up a user, friending 2 other accounts, migrating to a new host and checking out what the friends receive? 14:45 < fikabot_> 💬 !I didn't try the migration step in a sandbox 14:45 < fikabot_> 💬 I tried setting up Friendica in a sandbox 14:45 < fikabot_> 💬 Honestly looking at the traunch of shit connections part of me wants to write a simple contact viewer and pick the dozen or two I recognize and ditch the others 14:45 < fikabot_> 💬 It may be a decision made for me if this doesn't work of course :) 14:47 < fikabot_> 💬 This may be all federated "contact" management anyway...looking at the URLs in some of these logs the domain of that user doesn't exist in my exported dataset 14:51 < fikabot_> 💬 So I don't know what this thing is doing at this point with respect to the import...just free-associating 14:53 < fikabot_> 💬 Unless you all think this is a bad idea I'm going to delete that user and try to create it again from the import file 15:07 < fikabot_> 💬 Actually I’m guessing deleting and retrying won’t work since it has to reserve that username on that server doesn’t it? 15:48 < fikabot_> 💬 Before you do anything irrevocable, you still have a chance to repeat the sandboxed experiment with the full scenario. I mean, installing even 2-3 instances of Friendica only takes what, like half an hour? 15:49 < fikabot_> 💬 This time with increased log level as well perhaps. 15:53 < fikabot_> 💬 I think I’m going to add a feature to my archive browser that lets me see who I interacted with most and use this to manually rebuild my list and thin the crap 15:59 < fikabot_> 💬 For the short term anyway 15:59 < fikabot_> 💬 Coming up with a test rig to stress test migration could be interesting thoug 16:17 < fikabot_> 💬 Good thing you didn't start migration on the last day before the deadline! 16:17 < fikabot_> 💬 I moved my account once and I had mixed results. The contacts were notified and I received messages. However, the contact images pointed to the old nonexistent server and were not changed. I ended up creating a new account and contacting the old friends by hand. But that was around 3 years ago. 16:29 < fikabot_> 💬 > <@bkil:grin.hu> Good thing you didn't start migration on the last day before the deadline! 16:29 < fikabot_> 💬 16:30 < fikabot_> 💬 Well since it didn't seem to work anyway not sure that matters :) 16:31 < fikabot_> 💬 Although admittedly a mad rush with hours until a server is shut down would have been more stress inducing --- Log closed mar ene 18 00:00:39 2022