simonv3 has quit [Quit: Connection closed for inactivity]
heliostatic_ has joined #sandstorm
home has quit [Ping timeout: 272 seconds]
heliostatic_ has quit [Ping timeout: 264 seconds]
jdenz has joined #sandstorm
<jdenz>
Hi all. Back again with another question. :)
<jdenz>
This time it's about the git apps GitLab and Gogs. When attempting to clone or push a ~200 MB repository, I'm getting "fatal: The remote end hung up unexpectedly" errors. Has anyone else had this problem and is there an easy fix?
mnutt has joined #sandstorm
<kentonv>
jdenz: is this to a self-hosted server? If so, is it behind nginx?
<jdenz>
It's self-hosted and not behind nginx.
<kentonv>
hmm
<kentonv>
nginx disallows large uploads by default but I guess that's not it then
<jdenz>
searching on the net shows that I may need to change "GIT_HTTP_MAX_REQUEST_BUFFER"
<kentonv>
hmm, I guess that requires updating the apps
<kentonv>
I guess you should file bugs against them
<jdenz>
Ok, thanks. I figured it was likely app-specific. Is it complicated to change this on my local instance first, and then file the bug if it works?
<jdenz>
With both Gogs and GitWeb, I was able to successfully push the repository. However, with both, I'm unable to pull or clone it again. With GitLab, I wasn't even able to push it. All of the errors were "The remote end hung up unexpectedly." This makes me think there's a setting on my server I might be able to change.
<jdenz>
The errors in the Sandstorm debug logs are all similar to "** SANDSTORM SUPERVISOR: App exited due to signal 9 (Killed)."
<kentonv>
jdenz: that's weird!
<kentonv>
Sandstorm would not send signal 9 to the app and then report it in the log like that
<kentonv>
so apparently the app is killing itself? But that's weird because the app's root process should be sandstorm-http-bridge, and I don't know why the app would kill it...
<kentonv>
is there something on your system that might be going around killing processes?
<kentonv>
maybe they're being OOM-killed?
<jdenz>
Not that I know of. I don't know what OOM-killed is. :)
<kentonv>
out-of-memory
<kentonv>
the kernel will start killing processes when it's out of memory
<kentonv>
I wonder if the http-bridge process is somehow using a lot of RAM
<kentonv>
it shouldn't be buffering, though...
<jdenz>
Oh, yeah, that could be. At one point during my testing, I was getting some type of memory error, something about not being able to allocate it.
<kentonv>
maybe watch it under "top" and see if the process is eating memory
<jdenz>
Ok, I'll try that.
<kentonv>
shift+m to sort by memory usage
<kentonv>
(in top)
<jdenz>
The memory seems to be stable at between 34% - 45%.
heliostatic_ has joined #sandstorm
<kentonv>
for sandstorm-http-bridge?
<kentonv>
or for the whole system?
<jdenz>
sandstorm-http-bridge
<kentonv>
sandstorm-http-bridge shouldn't be using more than a few megabytes...
<kentonv>
so something is very wrong here
<kentonv>
that does seem to explain a thing or two
<jdenz>
oh, wait, that may have not have been sandstorm-http-bridge, sorry.
<jdenz>
Ok, I watched it closer and it ranges from not on the list (I guess 0) to 51%.
heliostatic_ has quit [Ping timeout: 244 seconds]
<jdenz>
After the pull process fails, it goes off the list again
<kentonv>
yeah so it sounds like there's some problem when processing large uploads and the process allocates a lot of memory and then gets OOM-killed
<kentonv>
it's weird because sandstorm-http-bridge is supposed to stream uploads.
<kentonv>
but it sounds like it may be buffering instead
<kentonv>
how much RAM do you have in this machine?
<ocdtrekkie>
FYI zarvox, your libsqlite3-dev suggestion got that build to work.
heliostatic_ has quit [Ping timeout: 250 seconds]
<paulproteus>
Yay
<zarvox>
ocdtrekkie: hooray! \o/
mnutt has quit [Quit: mnutt]
todayman has quit [Remote host closed the connection]
TC01 has quit [Remote host closed the connection]
shachaf has quit [Remote host closed the connection]
heliostatic_ has joined #sandstorm
micahd has joined #sandstorm
heliostatic_ has quit [Ping timeout: 240 seconds]
itscassa|away is now known as itscassa
jadewang has quit [Remote host closed the connection]
shachaf has joined #sandstorm
tobald has joined #sandstorm
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 240 seconds]
heliostatic_ has joined #sandstorm
heliostatic_ has quit [Ping timeout: 240 seconds]
tobald has quit [Quit: Ex-Chat]
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 250 seconds]
larjona has joined #sandstorm
heliostatic_ has joined #sandstorm
heliostatic_ has quit [Ping timeout: 260 seconds]
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 265 seconds]
michael_s has joined #sandstorm
itscassa is now known as itscassa|away
aggelos_ has joined #sandstorm
<aggelos_>
hello
<aggelos_>
tried using the install.sandstorm.io script but it seems to have failed when submitting the CSR to sandcats' /getcertificate
<aggelos_>
is that suposed to work?
<dwrensha>
yeah, you should be able to get a certificate
<dwrensha>
paulproteus is the sandcats expert. I don't think he's awake yet.
<aggelos_>
dwrensha: ok thanks. trying to do the curl invocation manually atm so hopefully I can get an error message
larjona has quit [Quit: Konversation terminated!]
<aggelos_>
nope, got the certificate manually
<aggelos_>
perhaps the 20 sec timeout is too small then?
<aggelos_>
I mean, it worked for me in <5s this time, but perhaps I was just unlucky when running the install scirpt?
<aggelos_>
hmm
<aggelos_>
now how to communicate to the script that it should continue and pick up the json file
home has joined #sandstorm
heliostatic_ has joined #sandstorm
home has quit [Ping timeout: 240 seconds]
heliostatic_ has quit [Client Quit]
<aggelos_>
well actually I'm in a bit of a rush so I restarted the process...
<aggelos_>
(oh I should mention the /getcertificate thing was a transient error)
<aggelos_>
mm that said the subdomain isn't resolving on chromium either
<aggelos_>
oh sorry my bad
<aggelos_>
resolves, just can't connect H)
<aggelos_>
*:)
<aggelos_>
hmm not totally clear whether email auth needs extra setup
jadewang has joined #sandstorm
<aggelos_>
I see some simplesmpt node thing, but I guess that needs a smarthost?
<aggelos_>
apparently yes
jadewang has quit [Ping timeout: 265 seconds]
<aggelos_>
shame, I should have read the docs more carefully. this isn't going to work for me atm :/ no google nor github account available
<aggelos_>
being able to create a user at the console would have been nice though
<dwrensha>
aggelos_: does the email login not work for you?
<aggelos_>
dwrensha: needs a smarthost which I don't have handy and it would take some time to set up... but I was able to recover the password for and old github account I had at work
<dwrensha>
I'm curious about this concept of a "smarthost". Would that be for incoming or for outgoing mail?
<aggelos_>
dwrensha: it's for outgoing mail
<dwrensha>
oh, so I suppose you could consider Mailgun and Sendgrid to be smarthosts
<aggelos_>
dwrensha: afaik smarthost is normally used in the context of configuration of an smtp server. those services seem to play the same role, yes
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 264 seconds]
TC01 has joined #sandstorm
jus has joined #sandstorm
NOTevil has joined #sandstorm
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 240 seconds]
pod has quit [Ping timeout: 265 seconds]
pod has joined #sandstorm
simonv3 has joined #sandstorm
mnutt has joined #sandstorm
jadewang has joined #sandstorm
jadewang has quit [Ping timeout: 255 seconds]
bb010g has quit [Quit: Connection closed for inactivity]
simonv3 has quit [Quit: Connection closed for inactivity]
<dwrensha>
which version of clang?
<zarvox>
he's also unhappy about the time involved in downloading Meteor and then Meteor downloading all its dependencies and throwing gigs of stuff in ~/.meteor
<zarvox>
dwrensha: unknown; I pinged him but he hasn't replied yet
<zarvox>
amusingly, my build tree appears to have used g++ to compile node-bignum, which suggests that if you export CC=clang CXX=clang++ you can't build Sandstorm
rchrd2 has joined #sandstorm
<zarvox>
reproduced on a clean clone with CC=clang CXX=clang++
<zarvox>
so clang-3.5.0 is definitely affected
<dwrensha>
where are you looking to see that g++ was used?
<warren>
crap. XMPP support is important to my own people so they can use OTR for person to person communication.
<zarvox>
warren: "Tell them they should use something more modern that implements the axolotl ratchet instead." More seriously: if they require end-to-end encryption, do they really care about who's providing the transport? Or is this to try to reduce the number of communications tools they need?
<warren>
the latter
<warren>
and I agree the desire to use OTR is incompatible with offline messaging
<zarvox>
Ahhh. And this needs to work with mobile as well? The worst things about OTR is no async session establishment, and terrible multi-device behavior
<warren>
right
<zarvox>
s/ is / are /
<warren>
It's a minority requirement, if it supported XMPP it would make him not concerned anymore
<kentonv>
I think we'll have an XMPP driver eventually.
<kentonv>
dunno what that means for rocket.chat, though
<zarvox>
Can you elaborate on the desired security properties of that requirement? Is it "confidentiality is resistant to service provider compromise" even though you'd likely run your own server anyway?
<zarvox>
or is it "confidentiality against service provider admins"?
<zarvox>
I guess the latter requires that the end devices hold the secrets in a way that the service and its scripts can't reach. Which maybe you could add to the desktop-app packages of rocket.chat?
home has quit [Quit: Leaving]
rustyrazorblade has quit [Quit: rustyrazorblade]
NOTevil has quit [Quit: Leaving]
jeffmendoza has quit [Ping timeout: 265 seconds]
jeffmendoza has joined #sandstorm
isd has joined #sandstorm
<maurer>
What's the general update time for sandcats.io domains? Hours? Days?
<zarvox>
maurer: are you asking "how quickly should DNS update when I change IPs?" The sandcats client attempts to send a UDP ping every minute.
<maurer>
zarvox: Yeah
<maurer>
Hm. Maybe I'm just getting hosed by ttl caches or something
<zarvox>
Possibly. What's your domain? I'll see what I'm resolving for it
<warren>
We're playing with wekan, we like it very much. It would be perfect if we had the ability to copy/move cards/lists between grains though.
<maurer>
(or I think I can, I haven't built sandstorm on this machine for a few months)
<paulproteus>
You'll "love" the most recent commit that fixes the crash.
<maurer>
Ah, wow, it was only a logging message :|
<paulproteus>
This is absolutely because I refactored something in a way that happened to be wrong but I wasn't testing it.
<zarvox>
Is this something we could have caught more easily with automated testing? Changing source IPs mid-test sounds somewhat challenging/invasive.
<maurer>
zarvox: I don't think this bug was caused by that, just noticed by it
<paulproteus>
I already have a test that triggers that in sandcats' integration tests.
<paulproteus>
But it does things on its own with Python, not with the shell.
<paulproteus>
Change an IP mid-test seems pretty doable. But even just _submitting_ the request would have found this issue.
<maurer>
Since the issue would only happen if there wasn't a csr, or the csr was malformed
<paulproteus>
The issue would happen for any request (i.e. "Update my IP address") that isn't a "getcertificate" call.
<maurer>
Ah, OK
<paulproteus>
And in fact, we could trigger that codepath more often (and therefore I'd see the crash on my own sandcats) by randomly updating the client's IP address.
<paulproteus>
The server can decide it's a no-op.
<paulproteus>
I had been thinking of doing that, but I just hadn't gotten around it.
<paulproteus>
BTW thank you _so much_ maurer for being "lucky enough" to hit this codepath.
<paulproteus>
len(None) would be a similar crash in Python, I suppose.
<paulproteus>
This is what I get for refactoring things! Oh well.
<paulproteus>
... I thought that I had written the above 4 lines in a different channel!
<paulproteus>
Oh well.
<paulproteus>
BTW maurer I have a build if you would like it. I'll link to it on the issue
<paulproteus>
maurer: I am grateful for your reports, and will wander off for about 45 minutes to a meeting. Seems like we will ship the fix in a day-ish.
<maurer>
paulproteus: Yeah, I'm waiting for the build to finish
<paulproteus>
Great
<maurer>
paulproteus: Would you mind making me a build? My build is breaking on an error where clang is upset with a weird idiom used by linux kernel headers
<maurer>
Hm. I wish gpg had a better web of trust searche4r, but it looks like you're only 2 jumps from people I personally know
* maurer
loads up tarball
jeffmendoza has quit [Ping timeout: 265 seconds]
larjona has quit [Quit: Konversation terminated!]
<maurer>
paulproteus: Can confirm, fix works
<maurer>
Aaand back on dev channel. Thanks, I'm going home
hedgehog has joined #sandstorm
<hedgehog>
Hi all, I'm putting together a capnp RPC demo app and having some problems figuring out how to pass capabilities from client to server using the Go bindings. Anybody have experience using the new zombiezen.com/go/capnproto2 library?
larjona_m has quit [Quit: AtomicIRC: The nuclear option.]
<kentonv>
hedgehog: Hi! Unfortunately I haven't tried them. Probably you should contact Ross (the maintainer) directly, or post to the capnproto mailing list.
<hedgehog>
Hey! Thanks. Actually maybe you can tell me if what I'm doing makes any sense. Briefly: the demo app is a chat server + client. The client connects to the server, uses interfaces exposed by the server to join a channel, and then joins the channel passing up a capability to allow the server to push messages back the other direction.
<hedgehog>
Conceptually should that work over the single connection opened by the client?
<hedgehog>
dwrensha: Thanks, good to know what I'm doing isn't obviously wrong :)
<hedgehog>
If this ends up looking like a good idea I'll clean it up and put it on GitHub, it should be simple enough to be a good example app for other people going down the same road.