<ocdtrekkie>
I just learned how to remove old kernels because my Ubuntu VM was way, way too full. And most of it was old kernels, it seems.
<kentonv>
ubuntu loves to allocate a tiny /boot and fill it with kernels
<zarvox>
Ubuntu still doesn't remove old kernels?
* ocdtrekkie
shrugs
<kentonv>
I have this problem all the time
<zarvox>
Huh. Fedora has removed outdated kernels for at least a few releases
<kentonv>
admittedly it's been a few releases since I updated
<ocdtrekkie>
My Ubuntu servers at my old job autoremoved kernels, but we also had a competent Linux user who set those servers up.
<kentonv>
(I don't use Ubuntu on my main machine)
<ocdtrekkie>
Unlike my Linux VM, which was set up by a Linux moron.
<kentonv>
well this is totally stupid behavior on Ubuntu's part, so don't feel bad. :)
<ocdtrekkie>
Anyways, I am going to follow the build instructions for the OC client hopefully before this weekend, so I'm hoping to have a server I can point it at so I can test mnutt's new toys.
<kentonv>
to test the owncloud client PR, what do you plan to use as the server?
<kentonv>
that is to say, I think you need to get a copy of mnutt's spk
<ocdtrekkie>
I have that already. He posted it... somewhere.
<ocdtrekkie>
Might've been IRC.
<zarvox>
"or package all of owncloud" :P
<kentonv>
how new? things changed significantly just before the PR was merged
<ocdtrekkie>
Random hint: If anyone has linked an SPK in a public place, I likely downloaded it. I will check with mnutt though if there's a newer one.
<zarvox>
hahahaha
<zarvox>
ocdtrekkie, our community SPK archival expert
<ocdtrekkie>
I am kinda assuming that janc wants :someone: other than the submitter to test that it functions, so I was like "okie dokie".
<ocdtrekkie>
You should see my pile of APKs from before Android DRM'd everything to heck.
<ocdtrekkie>
All sorts of old Android 2.x apps which still work and run faster and are smaller than... well, whatever gets shipped today.
<kentonv>
I tested the original etherpad spk again today
<kentonv>
still works
<zarvox>
backwards compatibility considered strong
<ocdtrekkie>
\o/
<ocdtrekkie>
I expect a lot of breakage once HackSession type stuff gets closed down though? For apps that use external resources.
<paulproteus>
06:20 < kentonv> I tested the original etherpad spk again today
<paulproteus>
06:20 < kentonv> still works
<paulproteus>
I love you kentonv
<paulproteus>
: D
<zarvox>
paulproteus: I love that your IRC client is on a server configured for UTC
<kentonv>
ocdtrekkie: we should be able to turn everything into powerbox requests, I think. UX might get annoying, but should still work.
<zarvox>
Yeah. Might have to do some grandfathering for the client-side sandboxing too.
<zarvox>
By which I mean "will definitely have to"
<kentonv>
ah, right, for client-side sandboxing my plan is basically that we'll have an API version bump, and apps declared to use the old API will get a big warning on install saying "this app will not be confined because it uses an old version of the API"
<kentonv>
most users don't _really_ care about confinement, so meh
<kentonv>
I suppose we could make it an advanced option to "confine this app anyway, I understand this may break it"
<zarvox>
Yeah, and that could be administratively enforced if e.g. some enterprise IT team wants to rely on Sandstorm's confinement
<paulproteus>
I'm an enterprising young chap
<paulproteus>
Which is just to say I'd probably turn that feature on if I could, damn the torpedoes, etc.
<zarvox>
Assuming that most apps that people want to use are maintained, unconfined apps probably aren't relevant to very many users, particularly now that we have update-checking
tobald has joined #sandstorm
<ocdtrekkie>
Makes sense.
<ocdtrekkie>
The big thing for me, is I don't particularly like being shoved up into an app version I dislike due to version breakage or a security hole. I'm currently on a vulnerable Android version because Google won't release the security patch to my device for my software version.
<ocdtrekkie>
I am kinda hoping Sandstorm taking care of most of the security aspects means I will not have to worry (as much) about such occurrences.
<paulproteus>
bd
<paulproteus>
I sometimes tell people, the main value of software is nostalgia.
<paulproteus>
Maybe this is because I'm still an emo teenager. But anyway, I am using the same window manager (awesome) as I started using in ~2006, and it's pretty seriously nice that I get to keep it.
<ocdtrekkie>
paulproteus : sometimes it's nostalgia. The reason cloud services and continuous development is horrible for seniors is that people move buttons a lot.
<ocdtrekkie>
(I will probably never recommend a cloud service to a senior. The webmail clients confuse them to no end as it is.)
<ocdtrekkie>
Sometimes it's just that the current generation of designers have very bad ideas lodged in their heads that need to be removed by power drill.
<ocdtrekkie>
I am actually unsure there is much software I use at all that's got an interface older than a few years.
neynah has joined #sandstorm
<ocdtrekkie>
I will emphasize particularly (hi neynah) that design changes on Sandstorm are done very well.
jacksingleton has quit [Ping timeout: 240 seconds]
jacksingleton has joined #sandstorm
jacksingleton has quit [Ping timeout: 240 seconds]
_whitelogger has quit [Remote host closed the connection]
_whitelogger_ has joined #sandstorm
cbaines has joined #sandstorm
ZucchiniZe is now known as ZucchiniZe-
ZucchiniZe- is now known as ZucchiniZe
ZucchiniZe is now known as ZucchiniZe-
_whitelogger has joined #sandstorm
todayman has quit [Ping timeout: 252 seconds]
svenneK has joined #sandstorm
<svenneK>
Hi, first of all thanks for a great project... I wonder about backups... the wiki says that you need to shut down sandstorm before backups...
<svenneK>
But if you have LVM snapshots avaible, that ought to work... Otherwise your software cannot be crash safe... (from the softwares perspective restarting from a LVM snapshot or a host crash should be the same)
<svenneK>
is it correct, that a LVM snapshot would also produce valid backups?
_iwc has joined #sandstorm
_iwc has quit [Killed (Sigyn (Spam is off topic on freenode.))]
<dwrensha>
hm. I suspect that the docs suggest shutting down Sandstorm first out of abundance of caution.
<dwrensha>
I don't know of any reason why it would be actually required.
_iwc has joined #sandstorm
larjona has quit [Quit: Konversation terminated!]
_iwc has quit [K-Lined]
<svenneK>
I am more concerned about the potentially lack of crash safety... actually shutting down sandstorm for backups is a minor problem
groxx has quit [Quit: Connection closed for inactivity]
<dwrensha>
I think at one point some folks raised some concerns about MongoDB's crash safety.
<dwrensha>
But I think that's a reputation problem for Mongo.
<dwrensha>
Like, long ago, there were issues but it's totally fine today.
<dwrensha>
Sandstorm should be robust to crashes. If it's not, that's a bug. I'm not aware of any way in which it is not.
<paulproteus>
I guess a difficulty right now is there's no straightforward way for Gogs-on-Sandstorm to be able to do the 'git fetch' step (though it could do that via HackSession's http download methods).
jadewang has quit [Ping timeout: 260 seconds]
<dwrensha>
zarvox: the specific thing I was trying to do was to add a "cancel" button to the accounts profile editor
<zarvox>
what would you cancel?
<dwrensha>
like, you've edited a few of the fields, then change your mind and instead only want to change a third field
<paulproteus>
about 4 times, until I felt I understood how "this" works.
<paulproteus>
Actually what I did was print it out, then sit on a chair and reread it, and then a week later, talk to someone who had more JS experience than I did, and attempt to repeat the content to them, and apparently impress them by showing them that I know how 'this' works which nearly no one does apparently.
paroneayea has quit [Read error: Connection reset by peer]
larjona_afk is now known as larjona
uppermgmt has quit [Ping timeout: 260 seconds]
<kentonv>
yeah, 'this' in javascript is horribly broken
<paulproteus>
larjona: Hmm, seems like maybe http://kanboard.net/ would be a useful Sandstorm package for Debconf purposes, if it solves the deadliens/dependencies aspect of kanban that madduck was concerned about?
<larjona>
paulproteus yesterday we were trying the demo. Next week we'll go on trying things, but frankly, I don't know if it solves his problems, and I don't know if the team would adopt it, even if it solves the problems (different people have different problems). And we have etherpad in storm.debian.net an people keep using titanpad.com...
<paulproteus>
I imagine they keep using titanpad.com because there's a very clear "Create a pad" page, and/or an easy URL for people to remember.
<larjona>
I will not give up in offering/explaining/trying to make it easy for people to use storm.debian.net, but the sysadmins already said that it was feasible to install kanboard in a Debian VM, so maybe people just will want to do like that (in the case they decide to use kanboard...)
<larjona>
my deb.li urls are very easy too :D
<paulproteus>
Cool (although I guess someone will have to figure out SSO for kanboard?)
<larjona>
no idea
<paulproteus>
Maybe it's worth asking a Titanpad user one-on-one why they pick Titanpad. My guess is they only skimmed your email.
<paulproteus>
I mean, a debconf team member who is creating Titanpad links.
<paulproteus>
Honestly, that's useful for _us_ (Sandstorm peeps) to know, what it takes for people to migrate!
* larjona
searches "skim" in the dictionary
uppermgmt has joined #sandstorm
<larjona>
yes, I'm bad at writing short, convincing emails
<paulproteus>
Hi uppermgmt, your domain name is terrifying but also great. (-:
<larjona>
I'll ask the titanpad fans why they use it
<rchrd2>
Speaking of javascript. Es6 has some nice new features. There are shorter function calls `myfunction= (arg) => arg + arg`. And this different function syntax keeps the same "this"
<zarvox>
impact is that dev packages are not overriding non-dev packages when resuming grains
<kentonv>
paulproteus: that doesn't have the mongo log
<kentonv>
zarvox: is this a regression?
<zarvox>
kentonv: yes :(
<kentonv>
paulproteus: it's definitely mongo that is dying. Unfortunately it likes to put any useful messages into mongo.log and print nothing useful on the console.
<paulproteus>
looking @ maongo log
<paulproteus>
+/- typos
<dwrensha>
MangoDB!
<kentonv>
zarvox: ok, well at least it doesn't affect oasis
<paulproteus>
2015-10-15T21:32:49.764+0000 [initandlisten] exception in initAndListen: 12596 old lock file, terminating
<paulproteus>
The official docs say if there's no replica sets, "*and* do not have journaling enabled"...
<dwrensha>
surely journaling is enabled?
<paulproteus>
So... presumably we should enable journaling?
<dwrensha>
the existence of /opt/sandstorm/var/mongo/journal seems to indicate that it is enabled
<paulproteus>
Beats me. The docs.mongodb.org link, to me, vaguely hints that if we had journaling enabled we wouldn't see this.
<paulproteus>
_Or_ maybe it means that it's safe to remove the lock file (?)
<kentonv>
we have journaling enabled, and we actually have replica sets enabled since that's the only way to get oplog tailing
<paulproteus>
Looks like if we start mongod with --repair always, then we are OK.
<kentonv>
but there's only one replica
<paulproteus>
But we would have to remove that lock file. "Use the --repair option. mongod will read the existing data files, write the existing data to new files and replace the existing, possibly corrupt, files with new files" "You must remove the mongod.lock file before using this procedure"
<kentonv>
btw I do not understand why the server monitor is aborting -- it should go on retrying starting mongo forever, I think
<paulproteus>
It seems to me what I should do is submit a Sandstorm pull request having us (1) check for this lock file, and if it exists, log a note saying "Aieee deleting the lockfile and starting mongod with --repair , hopefully your data is not spaghetti now" and (2) if that lock file existed, make sure to pass --repair into the mongod argv
<paulproteus>
We could decide to be more conservative and back up the current data before doing that, but it's not like there's anything else we would do with the old data.
<paulproteus>
Hypothetically if we screw that up by starting a mongod against this data directory when someone else decided to manually start a mongod against, then that'd be bad news.
<paulproteus>
We _could_ decide this is the time to introduce a 'sudo sandstorm repair' command, but I think our general perspective on CLI operations is "if it requires command line intervention, it doesn't exist".
<paulproteus>
I will get on that pull request, unless anyone else has any wiser ideas. If so, I'd be happy to hear them.
<paulproteus>
BTW pidfiles + process namespaces are a fascinating combination.
larjona_afk is now known as larjona
TC01 has joined #sandstorm
<kentonv>
paulproteus: what happens if you delete the pidfile, then the power goes out, and then it comes back up and you don't know to --repair?
<kentonv>
seems like you need some sort of dance here where you write some other file as a flag to tell you that repair is needed, and only delete it when mongo seems to have come up successfully
<zarvox>
do we even need pidfiles? I thought sandstorm supervised the processes
<kentonv>
worth checking also if maybe Mongo automatically deals with the lock file if you pass --repair; seems surprising that you'd have to delete it manually
home has quit [Quit: Leaving]
eternaleye has joined #sandstorm
jeffmendoza has quit [Ping timeout: 265 seconds]
jeffmendoza has joined #sandstorm
larjona has quit [Ping timeout: 255 seconds]
jeffmendoza has quit [Ping timeout: 265 seconds]
augustl_ has joined #sandstorm
zarvox_ has joined #sandstorm
todayman_ has joined #sandstorm
zarvox has quit [Write error: Connection reset by peer]