<shaiguitar>
dwradcliffe: is there any chance you can figure out the redis data part of that?
jalessio has joined #rubygems
<shaiguitar>
essentially there should be download count history in redis, but it seems old data is missing. was wondering if there were backups, or if we could figure out how to backfill it even? something to fix that inconsistency? (see link above)
huoxito has quit [Remote host closed the connection]
huoxito has joined #rubygems
<samkottler>
shaiguitar: we can't backfill it unfortunately
<shaiguitar>
evan: qrush drbrain https://github.com/rubygems/rubygems.org/pull/606 details the problem of missing download data reproduceable example with all the info of what's wrong. if there's something I can do to help let me know how...
<shaiguitar>
samkottler: is this a known issue?
<samkottler>
shaiguitar: it's not really an issue (backfilling) it's just a design thing
<shaiguitar>
samkottler: the totals coming back from the api are completely bogus compared to the sum it shows on rubygems.org
<samkottler>
shaiguitar: yeah I'm just reading the issue now
<shaiguitar>
samkottler: I'd be happy to help where I can, just don't have access to anything and don't want to step on any toes.
<shaiguitar>
evan: hey. yea :(
<samkottler>
evan: cmeiklejohn and I talked about maybe moving over to CRDT's for it
<shaiguitar>
I've been trying to figure this stuff out but it looks messy
<evan>
samkottler: it needs to be completely redone.
<samkottler>
evan: yep
arthurnn has joined #rubygems
<evan>
it's super dependent on redis atm.
<shaiguitar>
samkottler: yes, the test is not doing anything at all just proving a point
<evan>
i'm happy to use a CRDT for it.
<cmeiklejohn>
yeah
<cmeiklejohn>
+1
<cmeiklejohn>
i can't chat right now, but i've talked about this a few months back with sean cribbs
<cmeiklejohn>
and i think i have a solution for it
<shaiguitar>
I'm confused, is this related to #606 or #613 ?
<cmeiklejohn>
be back
<dwradcliffe>
606
<shaiguitar>
I'd like to help out with the counter stuff if I can, just not sure what I can do from where I am
<arohner>
samkottler: I have a (for now) reliable repro for the SSL problem. what's the way to get the most verbose SSL errors?
<shaiguitar>
also it'd be good to fix it going forward if there are problems with the current design, but is there any possible way we could backfill/migrate the missing data at all? make it consistent? I suppose there are 2 issues at play
<arohner>
I have ":verbose: :really" in my gemrc
<shaiguitar>
like I said, be glad to help if I can.
<samkottler>
shaiguitar: again, no, that data is collected at gem installation time
<arohner>
anything else that would be useful for that error?^^
<samkottler>
arohner: what are you trying to show? that gem requires a more recent rubygems
<samkottler>
"Try 'gem update --system' to update RubyGems itself."
<arohner>
than 2.1.9?
<shaiguitar>
samkottler: having the sum conform to what's returned from the api in necessary though. it doesn't have to be a true respresntation of what actually happened, but at least have the missing delta of any gem put back to the day it was built or whatever.
<arohner>
interesting, I just got OpenSSL::SSL::SSLError: SSL_connect returned=1 errno=0 state=SSLv3 read server hello A: wrong version number, from the aws sdk
<imperator>
arohner, did verbose output show anything?
<arohner>
this was during a unittest, not gem install
<dwradcliffe>
interesting
<imperator>
has its own cert file bundle
<imperator>
dunno if that's relevant
jnimety has quit [Quit: Computer has gone to sleep.]
dwknoxy has joined #rubygems
jnimety has joined #rubygems
jnimety has quit [Client Quit]
jfoy has quit [Quit: jfoy]
jnimety has joined #rubygems
vertis has joined #rubygems
havenwood has quit [Remote host closed the connection]
vertis has quit [Quit: vertis]
<headius>
hey guys
<headius>
I'm working on this bad_record_mac issue fetching gems with JRuby and starting to believe it's an rg.org issue
<headius>
we only started getting reports of this over the summer or so, and yet I can now reproduce it with versions of JRuby going back as far as last winter
<headius>
a couple reports online seem to indicate that there's buggy openssl versions out there that can cause this…I'm wondering about the rg.org service that handles download requests
<headius>
(apologize for plundering the dick gem thousands of times today)
<headius>
apologies, rather
<headius>
jruby 1.7.2 just reproduced it too, and I know we weren't getting reports of this back then
<headius>
at least one post related to my issue says it's a problem when server only supports SSLv3
<headius>
something in java secure socket stuff that (incorrectly?) picks TLS
<indirect>
yes
<indirect>
my issue contains a link to a Perl issue
stevenharman has joined #rubygems
<indirect>
that's a similar issue
<indirect>
client SSL version fallback
<indirect>
producing the wrong version
<headius>
you are getting it with MRI, right?
<headius>
my issue causes fail during connect handshaking randomly
<headius>
I have dumped SSL subsystem logs from JVM but they don't tell me a whole lot
<indirect>
headius: yes, I'm getting it all on MRI
<indirect>
if you could add your experience and repro case to my ticket, that would be awesome
<headius>
I was going to file another one, but I'll add to yours for now
<headius>
we've been getting it for many weeks or months but not before summer
<indirect>
yeah
<headius>
tell you what, I'll add an issue but reference yours too
<indirect>
I've been seeing slowly building reports for weeks
<indirect>
sounds good!
<indirect>
I strongly believe it's the same SSL issue with rg.org
<indirect>
but I have no objection to hopefully getting more attention :)
TheComrade has quit [Quit: TheComrade]
SilverKey has joined #rubygems
jkline has quit [Quit: jkline]
<headius>
yeah, if it ends up being a dupe so be it, but it's different enough I want to file it
<indirect>
headius: understood. thanks for reporting! I am hoping that someone will have a brilliant idea that solves it if we throw enough information at it :P
KayJails has joined #rubygems
<headius>
yeah we're getting a lot of reports about it lately :-(
<indirect>
headius: Travis has been seeing _tons_ of reports of it as well
<indirect>
I have been seeing it about once per 150 gem installs in Bundler builds on travis
<indirect>
which is about once per commit, since the bundler matrix is so big
<headius>
just to confirm, that's also still with MRI?
<indirect>
yes, across mri from 1.8.7 up through 2.1-dev
<headius>
I've been beating my head against this thing for a couple weeks now
<indirect>
and with rubygems versions from 1.3.6 up to 2.2-dev
<headius>
wow, that's a really happy bit of news, in a sad way
<indirect>
don't have any concrete patterns
<indirect>
on versions
<indirect>
but I haven't exhaustively researched it yet
<indirect>
they may be hiding in the data :/
<headius>
yeah
workmad3 has quit [Read error: Connection reset by peer]
SilverKey has quit [Remote host closed the connection]