#ceph IRC Log

Index

IRC Log for 2011-07-19

Timestamps are in GMT/BST.

[0:26] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) has joined #ceph
[1:16] * yoshi (~yoshi@KD027091032046.ppp-bb.dion.ne.jp) Quit (Remote host closed the connection)
[1:45] * Juul (~Juul@port80.ds1-vo.adsl.cybercity.dk) Quit (Quit: Leaving)
[1:47] * yoshi (~yoshi@KD027091032046.ppp-bb.dion.ne.jp) has joined #ceph
[1:47] * yoshi (~yoshi@KD027091032046.ppp-bb.dion.ne.jp) Quit (Remote host closed the connection)
[2:03] * Tv (~Tv|work@ip-64-111-111-107.dreamhost.com) Quit (Ping timeout: 480 seconds)
[2:31] * joshd (~joshd@ip-64-111-111-107.dreamhost.com) Quit (Quit: Leaving.)
[2:46] * phil__ (~quassel@chello080109010223.16.14.vie.surfer.at) Quit (Remote host closed the connection)
[2:54] * cmccabe (~cmccabe@69.170.166.146) has left #ceph
[3:47] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) Quit (Quit: This computer has gone to sleep)
[4:43] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) Quit (Read error: Operation timed out)
[5:35] * yoshi (~yoshi@p4094-ipngn1601marunouchi.tokyo.ocn.ne.jp) has joined #ceph
[7:15] * yoshi (~yoshi@p4094-ipngn1601marunouchi.tokyo.ocn.ne.jp) Quit (Remote host closed the connection)
[8:29] * yoshi (~yoshi@p4094-ipngn1601marunouchi.tokyo.ocn.ne.jp) has joined #ceph
[10:26] * Dantman (~dantman@S0106001731dfdb56.vs.shawcable.net) has joined #ceph
[10:40] * masACC is now known as maswan
[12:53] * yoshi (~yoshi@p4094-ipngn1601marunouchi.tokyo.ocn.ne.jp) Quit (Remote host closed the connection)
[14:26] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) has joined #ceph
[14:30] * ghaskins_mobile (~ghaskins_@66-189-113-47.dhcp.oxfr.ma.charter.com) Quit ()
[14:36] * aliguori (~anthony@cpe-70-123-132-139.austin.res.rr.com) has joined #ceph
[15:12] * darkfaded (~floh@188.40.175.2) has joined #ceph
[15:14] * darkfader (~floh@188.40.175.2) Quit (Remote host closed the connection)
[16:43] * Tv (~Tv|work@ip-64-111-111-107.dreamhost.com) has joined #ceph
[16:48] * greglap (~Adium@166.205.138.14) has joined #ceph
[17:23] * Dantman (~dantman@S0106001731dfdb56.vs.shawcable.net) Quit (Remote host closed the connection)
[17:41] * greglap (~Adium@166.205.138.14) Quit (Quit: Leaving.)
[17:41] * foxhunt (~richard@109.109.115.145) Quit (Remote host closed the connection)
[17:46] * bchrisman (~Adium@c-98-207-207-62.hsd1.ca.comcast.net) Quit (Quit: Leaving.)
[17:58] * jbd (~jbd@ks305592.kimsufi.com) has left #ceph
[18:38] * bchrisman (~Adium@70-35-37-146.static.wiline.com) has joined #ceph
[18:46] * joshd (~joshd@ip-64-111-111-107.dreamhost.com) has joined #ceph
[18:55] * jbd (~jbd@ks305592.kimsufi.com) has joined #ceph
[18:55] * cmccabe (~cmccabe@69.170.166.146) has joined #ceph
[18:57] * wonko_be_ (bernard@november.openminds.be) has joined #ceph
[18:57] * wonko_be (bernard@november.openminds.be) Quit (Read error: Connection reset by peer)
[19:07] <gregaf> bchrisman: sage said that you had a binary that was revealing the bug with wait locking back before it got fixed
[19:07] <gregaf> any chance I could get my hands on the source?
[19:08] <bchrisman> yeah??? there's a source prog
[19:08] <bchrisman> Last I checked, it was working
[19:09] <bchrisman> I'll track it down
[19:09] <gregaf> thanks
[19:09] <gregaf> we want to get it or something like it into our QA :)
[19:09] <bchrisman> you guys seeing problems like that?
[19:09] <bchrisman> ahh ok??? yeah.. that's great.
[19:09] <gregaf> there aren't a lot of programs available for testing file locking
[19:09] <gregaf> there's a locktest in xfstests that does a fair bit
[19:10] <gregaf> and I hacked it up some so it could test wait locking, but due to the way it works testing for non-compliant behavior with wait locks would be???difficult ;)
[19:34] <sagewk> general question on the proflogger unix domain socket. do we want to make a general purpose socket for communicating with the daemon, that takes a command, of which "dump proflogger counters" (or whatever) is one command? or just stick with the current socket that's created specifically for the instrumentation...
[19:40] <cmccabe> I guess taking a command would be more flexible
[19:41] <cmccabe> we already do have the injectargs interface though
[19:41] <sagewk> that's a one-way channel that requires authentication with teh monitor and a circuitous route across the network
[19:42] <cmccabe> true
[19:42] <sagewk> the idea here is things like the proflogger stats on the local node.. or asking the daemon to shut down.. or a "are you alive" check
[19:43] <cmccabe> I guess the client could send 4 bytes with a command code or something
[19:43] <cmccabe> which could always be 1 for now
[19:43] <cmccabe> I don't see a really compelling reason _not_ to do it
[19:43] <sagewk> this, incidentally, takes the prologger uri thing i didn't much like out of the equation, since the domain socket interface is no longer in scope for the proflogger itself
[19:43] <cmccabe> I don't understand
[19:44] <cmccabe> the socket needs to be somewhere, there still needs to be a way of configuring that
[19:44] <sagewk> yeah
[19:44] <cmccabe> I think one day someone may want a TCP interface
[19:44] <cmccabe> it's not that unreasonable
[19:44] <cmccabe> and easy to support with the URI mechanism
[19:45] <sagewk> but the worry about user confusion with the domain socket configuration vs enabling/disabling the instrumetnation itself goes away since they are not directly related any more
[19:45] <cmccabe> I don't think we want to be creating sockets that will never be used
[19:45] <cmccabe> if nothing else, TV will howl about how it messes with the unit tests
[19:45] <cmccabe> and crushtool, etc etc
[19:45] <sagewk> and its just as easy to have an option for each interface method instead of list of uris
[19:46] <cmccabe> it's just harder on the actual user
[19:46] <sagewk> only daemons would ever enable such an interface
[19:46] <cmccabe> who will have to figure out why he configured socket_path and tcp_setup and it still doesn't work
[19:46] <cmccabe> because there's some 3rd variable that's on/off
[19:47] <cmccabe> I guess it's less of a concern than it was for logging, because users don't have the same level of expectations
[19:47] <sagewk> i think that's a moot point; both would be on by default, and they are unrealted (a communication channel to the daemon vs profiling counters)
[19:48] <cmccabe> they can be on by default for daemons, but I don't think they should be for other software
[19:48] <sagewk> right
[19:48] <sagewk> definitely not :)
[19:48] <cmccabe> and even if they are on by default, I can't think of a really good default socket location
[19:48] <cmccabe> maybe you have one in mind in /var?
[19:48] <sagewk> something like /var/run/ceph/$name.socket
[19:54] <sagewk> my main hangup with the uri list is just that we don't use uris for any other configuration anywhere else in the project. separating the communication channel from the instrumentation itself seems simple and straightforward and seems to avoid the issues that were worrying you before
[19:58] <cmccabe> I still would like to rework the logging configuration so there's just one option, which takes a list of URIs
[19:58] <cmccabe> because I still get questions about why --log-to-stderr doesn't disable logging to files, and why --log-file doesn't disable logging to syslog, etc
[19:59] <cmccabe> even from people on our own team
[19:59] <cmccabe> so it really seems like a bad interface, or at least one that surprises people
[20:11] <cmccabe> I don't feel super-strongly about it for proflogger, though. I think people don't have the same expectations for an interface like this.
[20:11] <cmccabe> so having separate configuration variables for unix domain sockets vs. tcp, etc. could be doable.
[20:12] <cmccabe> but even then you run into awkward situations like people configuring the TCP interface, and being confused that the UNIX domain socket is still being created... because it's in the defaults.
[20:14] <cmccabe> also, I remember you were thinking about different names. How do you like PerfCounter?
[20:28] <Tv> cmccabe: hey Colin do you have a moment?
[20:28] <cmccabe> tv: yeah
[20:28] <Tv> jabbering..
[20:30] * mtk (~mtk@ool-182c8e6c.dyn.optonline.net) Quit (Quit: Leaving)
[20:34] * jbd (~jbd@ks305592.kimsufi.com) has left #ceph
[21:28] * cmccabe (~cmccabe@69.170.166.146) has left #ceph
[21:55] * cmccabe (~cmccabe@69.170.166.146) has joined #ceph
[22:07] * cmccabe (~cmccabe@69.170.166.146) has left #ceph
[22:41] * cmccabe (~cmccabe@69.170.166.146) has joined #ceph

These logs were automatically created by CephLogBot on irc.oftc.net using the Java IRC LogBot.