[00:12:29] --- manfred furuholmen has become available [00:33:12] --- dev-zero@jabber.org has become available [00:57:17] --- Claudio Bisegni has become available [02:01:39] --- dwbotsch has left [02:17:31] --- dwbotsch has become available [02:37:23] --- Claudio Bisegni has left [03:05:26] --- Simon Wilkinson has become available [03:12:33] --- Simon Wilkinson has left [03:14:46] --- Simon Wilkinson has become available [03:30:02] --- manfred furuholmen has left [03:38:44] --- Simon Wilkinson has left [04:25:18] --- Simon Wilkinson has become available [04:30:56] --- Simon Wilkinson has left [04:33:51] --- Simon Wilkinson has become available [04:50:48] --- Simon Wilkinson has left [04:54:36] i'm sad i failed math. anyway, none of it matters. no directories were harmed in my failure to do math. but we really need to kill those directories anyway. [04:56:21] hm. it would be even cooler if all my jabber clients were happy [04:57:03] --- Simon Wilkinson has become available [05:05:50] --- Simon Wilkinson has left [05:31:15] and then lars replied to mail an hour ago with different wrong numbers [05:46:51] --- Simon Wilkinson has become available [05:49:41] --- tkeiser@sinenomine.net/owl has left [05:56:05] --- matt has become available [06:08:09] --- Simon Wilkinson has left [06:19:19] --- Derrick Brashear has left [06:19:34] --- Derrick Brashear has become available [06:54:20] --- dmontuori has become available [06:57:34] --- Derrick Brashear has left [07:00:12] --- Derrick Brashear has become available [07:00:44] --- haba has become available [07:04:05] --- Derrick Brashear has left [07:06:48] --- Derrick Brashear has become available [07:09:15] --- Simon Wilkinson has become available [07:10:19] Hi Simon. I head you were in the theater business yesterday when I had an openssh question for you. [07:10:34] --- Derrick Brashear has left [07:10:44] --- Derrick Brashear has become available [07:10:49] Indeed I am. Just opened a show in Cambridge. [07:12:04] --- Manfred has become available [07:12:22] --- Simon Wilkinson has left [07:13:25] Ah, nice. The question was if you know if modern openssh with your patches still logs logins as two rows, one for the gssapi success and one for the computer the login came from. It would be much nicer to have only one row with all information for log processing. [07:13:48] Eh, I don't type fast enough :-| [07:15:03] his network sucks [07:15:05] he'll be back [07:15:25] --- Derrick Brashear has left [07:15:37] My network sucked today too (had to repair it all morning) [07:24:01] haba: where can I find the afs interface in the heimdal code (kafs) ? [07:27:27] libkafs? [07:27:44] the afslog program? [07:28:28] I guess the afslog program is the example how to use libkafs. [07:30:07] perfect [07:31:01] Wait with that until you have read the code :) [07:31:19] hehe it is not so simple [07:31:44] It is functional programming brought to C. [07:32:14] somthing like hdb [07:33:04] Anyway i'd like to change the samba implementation on the ticket creation .. with less license problem .. [07:37:22] --- Derrick Brashear has become available [08:01:36] --- Simon Wilkinson has become available [08:02:28] haba: I'll take a look at fixing that. [08:03:10] Simon: That would be great and the folks analyzing the logs for "strange logins" would be very happy. [08:04:40] I think you'll still end up with two log lines, but I can make the GSSAPI one have more information. [08:05:36] I think the most important stuff that folks want together is the principal name and the IP addr. [08:06:19] Okay. We should be able to do that. [08:06:46] I can't promise it will ever go into "vanilla" OpenSSH, but I'll certainly make the change for the next version of my patchset. [08:07:00] that's good enough for me [08:07:02] :) [08:10:31] Derrick (openafs-info): Muahaha, yes, you can understand me wrong if you want. [08:12:11] I'm sure Derrick would never misunderstand something for comic effect. [08:12:15] well, we were talking about solaris, and you started talking about some other thing [08:12:25] it wasn't even for comic effect [08:12:51] I only topicdrifted a little towards tmpfs [08:12:59] but it's not even the same tmpfs [08:15:26] I's a filesystem on which normal diskcache is not happy. Other than that sure different. [08:16:28] well, try the same ifdef i told doug about, new module and afsd, and see if you have any luck. [08:16:31] I think there may still be some locking issues on Linux. If you take a look at the RT tickets for Marc's changes which added open by path for Linux, you'll see more of the back story. [08:16:41] marc dionne believes there may yet be locking issues. i suspect he's correct. [08:17:04] if you do deadlock it, get a backtrace from alt-sysrq-t, and maybe we can fix it [08:39:03] --- Simon Wilkinson has left [08:39:56] --- Simon Wilkinson has become available [08:48:33] time to head home soon. [08:48:50] --- haba has left [08:49:20] Derrick - I've been seeing some weird "hesitations" on one of the two computers that kernel panics w. afs every so often [08:49:27] I've tracked it down to afs [08:49:54] doing a tcp dump, what I'm seeing is a bunch of "FS Request: fetch-status" going to the afs server, and about every 3 of those an RX ABORT comes back [08:50:21] this goes on for a good while [08:51:03] brb... goddamn jabber [08:51:06] --- dwbotsch has left [08:51:25] --- Rrrrred has become available [08:51:53] I'm gonna venture a guess that this is somehow related to the kernel panics, but even if not related, still a problem that needs to be solved [08:52:36] yeah, you're hitting the fileserver throttling on unreadable stuff. i doubt it's related. if you want to solve it, well, your options have been discussed at length on the list before [08:53:14] well... once the client has things cached, it's fine [08:53:27] what do you mean by "unreadable" ? [08:53:28] dump with tcpdump's -x, get a full payload, and decode the last 32 bits of the abort. it will be a translate_et able error. i bet it's UAEACCES [08:53:36] i mean "you can't read it" [08:54:03] --- dmontuori has left [08:54:09] --- dmontuori has become available [08:55:47] fs reply fetch-status error #49733388 (32) [08:56:45] #define UAEACCES (49733388L) [08:56:49] unreadable. [08:56:54] to you [08:57:06] (well, to whatever is trying to access it) [08:57:19] you know that error code off the top of your head? impressive [08:57:49] well, i confirmed it, hence the output. but this is a routine problem. [08:58:05] except I have permissions to read everything [08:58:08] transarc added "don't hammer the fileserver" delay logic. i think it's.... less than fortunate? [08:58:19] well, what's the request which is failing? [08:58:22] this happens, for example, when I'm vimming a file in my homedir [08:58:33] like, what was fetch-status'd? [08:59:09] can that be found in the rx packet? [08:59:18] I've got an fid [08:59:28] 536870930/892/1836 [09:00:20] ok. well, use ls -li and find inode 1180540 [09:00:28] you may need to recurse [09:00:52] offhand guess: it's access being done as not you, it's not system:anyuser accessible, and something no longer has authentication [09:01:16] well... the processes that end up in disk wait are running as me [09:01:30] java, metacity, aspell [09:02:11] how did you get 1180540 out of 536870930/892/1836 ? [09:02:35] /afs/andrew.cmu.edu/usr/shadow/calcinode.c [09:02:41] not mine, but i keep a copy around [09:02:46] see also calcfid.c [09:05:06] in /vicepa ... find . -type d -exec ls -li {} \; |grep 1180540 returns nada [09:05:35] in your homedir [09:05:39] not in the vice partition [09:05:44] oh, in my afs homedir ? [09:05:52] the inode numbers in the vice partition are totoally not relevant [09:06:06] does it matter if I do that on the same client or a different client? [09:06:06] also, finding the file there would tell us *nothing* [09:06:14] should be same everywhere [09:09:13] do we know that it's actually looking for that inode in my volume? [09:10:14] The fid tells you [09:10:23] the fid says it's whatever volume 536870930 is. you tell me [09:10:42] yep, that's me [09:10:50] ok then [09:10:53] so i guess we know [09:15:49] so, 892 is the vnode number... what's the 1836 ? [09:16:10] uniq [09:16:26] a generation number, effectively. not input to the inode number [09:16:41] almost sounds like a versioning thing [09:18:34] yes. it is. [09:18:44] but it has nothing to do with your problem [09:22:16] hmmm... this could take a while [09:24:11] --- agoode has left [09:50:03] --- Russ has become available [09:54:25] --- Simon Wilkinson has left [10:17:27] hmm... the offending file in this case, is ~/.recently-used [10:17:33] which I certainly can access [10:20:39] same file for the completely different app running java [10:21:50] what's the acl on the directory? [10:22:17] like, i assume what's happening is the application somehow loses authentication, and thus... [10:22:52] cnfhosts l grp_all l system:administrators rlidwka admin rlidwka dwb rlidwka dwb7@cit.cornell.edu rlidwka [10:23:13] my whole login session should be inside the pagsh w. my tokens [10:23:55] I have yet to figure out what app is trying to get to that file -- lsof is no help [10:24:08] Wait, this is starting to sound vaguely familiar. Can someone bring me up to speed? [10:24:14] yeah. so without tokens it'd lose [10:24:25] except I certainly do have tokens [10:24:37] I even have a krenew running to ensure that I keep having tokens [10:25:02] and the process was started in the pag you're krenewing in? [10:25:17] if so, it suggests perhaps the issue is background daemons storing [10:25:19] well, the various apps (vim, java) are [10:26:42] jhutz - scroll up :) [10:28:44] --- Simon Wilkinson has become available [10:30:48] no [10:30:55] no what? [10:32:58] No, I won't scroll up. I just don't have time today to catch up on this the hard way. If you can't or don't want to summarize, that's OK; the chance that I can offer anything useful on this one are minimal anyway [10:32:59] Derrick - different app, now same thing happening w. a different inode [10:33:47] jhutz - apps freezing, seeing fetch-status on fids and a return of EUACCES [10:34:00] i can only assume they're doing unauth connections. rxdebug will tell you, since tcpdump gives you the epoch and cid, you can rxdebug and see if it's an authentic connection or not [10:34:08] if not, well, the question is why, but that's the issue [10:34:20] apps freezing is not the issue [10:34:33] apps freezing is the symptom [10:34:51] derrick - I only see a fid in the tcpdump output [10:35:07] like, unauth accesses return errors. the server throttles your access. i don't actually care about it, because it's not the issue. if you want to bark up that tree, you will never get an answer, because it's not the interesting question [10:35:36] use -vv? [10:36:12] ok, see the cid [10:36:48] don't see anything labeled epoch [10:37:04] the cid is enough. look for that cid in the rxdebug output [10:37:07] --- Simon Wilkinson has left [10:37:28] rxdebug on the client? [10:37:36] --- Simon Wilkinson has become available [10:37:44] on the server is more interesting [10:38:02] fileserver port? [10:39:11] rxdebug server not finding that cid [10:39:15] yes [10:39:18] on the client port 7001 it is [10:39:37] tcpdump also showing a bunch of [bad udp cksum !] in the output [10:40:05] sorry, i meant against the server's server port. but the client's client port, what do you see? [10:41:29] Connection from host 128.253.198.27, port 7000, Cuid bde190ab/24432f14 serial 1131, natMTU 1444, security index 0, client conn call 0: # 408, state active, mode: receiving, flags: reader_wait, has_output _packets call 1: # 0, state not initialized call 2: # 0, state not initialized call 3: # 0, state not initialized [10:42:23] that's rxdebug client 7001 [10:43:45] --- Simon Wilkinson has left [10:43:46] well, this is simple [10:43:52] good [10:44:30] > tcpdump also showing a bunch of [bad udp cksum !] in the output you're running tcpdump on one of the machines involved? many systems have the behavior that packet capture tools don't see the "real" checksum on outgoing UDP packets, and so think every outgoing packet has a bad checksum. You can generally ignore that. [10:44:30] are you 128.253.198.110? [10:44:37] yep [10:45:08] that's the client [10:45:12] Connection from host 128.253.198.110, port 7001, Cuid bde190ab/24432f14 serial 473, natMTU 1444, security index 0, server conn call 0: # 465, state dally, mode: error [10:45:31] (rxdebug's -allconn flag) [10:45:44] ah, ok [10:46:28] note, though: security index 0 [10:46:57] ok [10:47:32] further, i'll guess your tokens expire in 5.8 hours [10:47:56] can't tell you for sure since I can't get to a terminal to type "tokens" at the moment [10:48:09] --- dmontuori has left [10:48:10] but they are 8 hour tokens that are renewed by krenew or whenever I unlock the screensaver [10:48:15] --- dmontuori has become available [10:48:28] ah. perhaps not. well, there are at least 3 authenticated sessions from that host. i can't tell whose, but... [10:48:43] many other folk are logged into that host as well [10:48:47] we *may* be able to tell why the fetchstatus is not authentic from fstrace output [10:49:01] but the fundamental issue is "it's not an authentic request" [10:49:22] that's indicated by the security index 0 ? [10:49:34] rxnull. [10:49:39] kad is security index 2 [10:49:49] e.g. Connection from host 128.253.198.110, port 7001, Cuid bde190ab/32c8694 serial 2, natMTU 1444, flags pktCksum, security index 2, server conn rxkad: level clear, flags authenticated pktCksum, expires in 5.8 hours [10:50:09] rxkad is set to clear on that host, I believe [10:50:22] yes, that's fine [10:50:28] it's still security index 2 [10:50:44] ok, so, fstrace [10:50:50] fstrace -clear, fstrace -on ? [10:51:13] fstrace setset cm -active [10:51:16] wait for problems [10:51:19] fstrace dump cm [10:51:31] k... problems are happening as I type [10:52:02] if you don't have /usr/vice/etc/C/afszcm.cat (or equivalent path) installed, the result will be useless. if every line is "raw op "something, don't bother. i can't read those [10:52:28] I see no such file... must not be part of the rhel4 rpms [10:53:16] well, you can try. we'll see. but if it's all raw ops, i can't read it [10:53:18] --- tkeiser@sinenomine.net/owl has become available [10:53:37] is it a file that can just be copied from somewhere else? Or is it unique to each compilation of openafs? [10:54:40] if you copy it from another intel box you'll be fine. it may be byte order dependent. [10:55:02] well... my only intel/amd boxes are the rhel4 boxes... would you happen to have one? [10:55:16] I have some rhel5 ones [10:55:34] (I'm trying to think where I might find a rhel4 one..) [10:55:46] the one in /afs/andrew/usr/shadow should work [10:55:49] hmm... there seems to be such a file in the source tree [10:56:01] ok, so copy it to said location... do I then need to restart something for it to be seen by fstrace? [10:56:07] nope [10:57:03] I have a rhel4 one if needed. [10:58:36] stevenjenkins - stick it someplace? [10:58:41] the file, that is? :) [11:00:44] sure. one sec. [11:03:24] try http://steven.endpoint.com/afszcm.cat.rhel4 [11:04:33] awesome... now we're in business [11:05:44] --- dev-zero@jabber.org has left [11:07:41] --- Manfred has left [11:10:55] Derrick - dump is in /afs/cnf.cornell.edu/shares/public/outside_users/oafs-fetchstatus [11:24:24] time to run down to the framing store [11:24:49] oooo, who are you going to frame? [11:25:20] roger rabbit. that bunny has it coming to him. [11:29:45] grrr... distracted by user [11:29:47] now to head down [11:32:21] users suck :-) [11:36:32] hm. pid 15263? [11:38:54] of course i see in the first dump exactly one error [11:45:12] yeah. exactly one error. [11:48:48] --- Simon Wilkinson has become available [11:50:33] If the afszcm.cat file isn't in our RPMs, please file a bug. [11:51:13] that too [11:51:37] (I can't easily look now...) [11:53:27] --- Simon Wilkinson has left [11:53:46] --- dmontuori has left [11:53:52] --- dmontuori has become available [12:24:34] --- reuteras has left [12:33:16] exactly one error? [12:34:15] simon - bug filed [12:34:55] look yourself. search for code 0x2f6df0c [12:35:22] ENOENT (code 2) doesn't count. code 0 is success. there are some code 99999, but there's exactly one EACCES there [12:36:11] ok [12:38:04] is it clear that that is even an EACCES having to do w. my particular issue? [12:39:13] it's the cause of the aborts you reported. it will trigger that particular delay. beyond that, no, nothing is clear [12:41:08] so, one EACCESS somehow causes a bunch of RX_ABORTS afterwards [12:45:00] it causes a number equal to the number of retransmits the client did. it's not some random number [12:45:11] well. [12:45:13] not really [12:45:25] a retransmit will be answered just once. [12:45:45] what are the call numbers of the aborts? [12:46:23] the call numbers are incrementing [12:47:09] there will be 3 fetch-statuses with a call number, then a fetch-status error #49733388 w. the same call # back [12:47:15] right now at 263 [12:47:42] ok, so that is clearly the throttling [12:47:54] it's retransmitting, and getting one answer for the 3 attempts [12:47:58] which is fine, and correct. [12:48:12] the question is why fstrace isn't showing more than one failure [12:48:28] 49733388 is 2f6df0c [12:48:43] so basically, for every such reply i want to see a line in fstrace which is that [12:48:57] I just did a few more fstrace dumps, and have, well, ONE error in each dump [12:49:33] of course the other thing is... 0x4112616f only appears isn the fstrace for those errors. nowhere else. [12:49:34] maybe I should make the log bigger? [12:49:44] does "groups" show a pag group, or do we have to deal with keyrings? [12:50:05] i want to know if PAG 1091723631 has tokens [12:50:20] groups shows pag groups [12:50:28] if that's the correct pag, then yes [12:51:12] waiting for groups to return [12:51:50] user 0x4112616f in the fstrace is a literal pag number. de-hex it, and in "onegroup env", that's the group you see. which is 1091723631 [12:51:59] pag is 1091723634 [12:52:11] yeah, well, that's not the one this request is coming from [12:52:16] you have something running in another pag [12:52:23] --- dmontuori has left [12:52:29] --- dmontuori has become available [12:52:41] ps auxwwg *may* tell you interesting things. iirc. maybe. [12:53:10] but the pag in question is 1091723631, not 1091723634. so something is in a different pag, which i bet has no tokens. [12:53:15] and that is the underlying issue [12:53:17] wait [12:53:28] using a login manager of some sort? xdm or that ilk? [12:53:32] gdm [12:53:56] which authenticates and authorizes via pam (which gets the tokens) [12:54:15] and was whateverpid 15263 is started from a gui menu? [12:54:40] 15263 is gam_server [12:54:49] I think that's started along with nautilus during the login [12:55:15] tho, until I can get it to screw up again in a program other than nautilus, it's unclear [12:55:32] right now, the java app and vim are behaving... nautilus is failing every time [12:55:36] well, this screams "not running as you, no permissions, just making you miserable" [12:55:42] and it probably is the problem [12:56:09] i assume it's the famd replacement, the thing which uses inotify to monitor for file changes, and it's crawling your homedir [12:56:22] since I just opened a file browser window, yes [12:56:30] wikipedia says i am right [12:56:52] ok. well, gam_server needs to be started from your session. [12:57:07] i know diddly about what you'd need to do to do that [12:57:10] it was [12:57:59] a gam server should have something to do with legs, if you ask me. [12:58:20] in fact, this has happened when I've been working, only in xterm, in a folder completely outside of my homedir [12:58:37] well, "this" [12:58:54] i assume you didn't have tcpdump running, and you are commenting on observed behavior [12:58:55] things going screwy. [12:58:58] yes [12:59:06] yes, the nebulous "it broke" bug report [12:59:10] yay! [12:59:10] now that I do have tcpdump and fstrace, I can certainly try and capture things [12:59:17] It worked before it broke! [12:59:17] is it time to drink yet? [12:59:23] yeah yeah [12:59:38] i had a huge argument with (redacted) about "it", "that" and "broken" [12:59:41] hmm... it's after noon in NY state, so, yes, it's time to drink [12:59:48] i'm in pa [12:59:56] where it is1 minute to 4 [13:00:28] tho, I think one can buy beer in the store starting at 11AM [13:01:49] you know, I wonder if gam_server sometimes hangs around from a previous login and then the new login connects back to it [13:02:50] possibly [13:03:07] it was actually still there after having loggedout [13:03:09] it's effectively being treated by the fileserver as DoSing the fileserver. [13:03:13] killed it and logged back in [13:03:23] things seem happy, n ow [13:04:04] so, how did you determine what pag that pid was running in? [13:04:23] i explained that. "scroll up" :) [13:04:36] grrrr :) [13:04:44] you reap what you sow [13:04:57] quoting myself: user 0x4112616f in the fstrace is a literal pag number. de-hex it, and in "onegroup env", that's the group you see. which is 1091723631 [13:05:16] like, the error in fstrace was time 404.055354, pid 15263: Analyze RPC op 2 conn 0xdfe0dd00 code 0x2f6df0c user 0x4112616f [13:05:17] --- agoode has become available [13:05:22] "user 0x4112616f" [13:06:21] coolness [13:09:14] you're right... this was simple :-D [13:10:00] i am often right. i am occasionally way off. neither does me any good [13:10:33] --- Derrick Brashear has left [13:18:58] but this may have done me good, if you are truly right [13:21:07] --- agoode has left [13:28:30] --- dmontuori has left [13:28:58] --- tkeiser@sinenomine.net/owl has left [13:30:03] --- agoode has become available [13:30:16] --- agoode has left [13:30:16] --- Adam Goode has become available [13:34:40] --- Derrick Brashear has become available [13:44:15] --- dmontuori has become available [14:04:51] --- Manfred has become available [14:09:58] --- tkeiser@sinenomine.net/owl has become available [14:29:36] --- Manfred has left [14:34:23] --- canehan has become available [14:50:54] --- canehan has left [16:13:02] --- Marc Dionne has become available [16:16:52] Howdy. cache on tmpfs works with the new exportfs code in 1.5.55 [16:18:36] I'm running a test right now with: tmpfs 102400 63284 39116 62% /var/vice/cache [16:43:36] tmpfs is exportable? cool! [16:45:50] yup, it supports the exportfs API [16:54:32] ... since 2.6.19 if I look at the code [18:22:51] --- Marc Dionne has left [18:35:07] --- jhutz@jis.mit.edu/owl has left: Lost connection [18:35:07] --- SecureEndpoints has left: Lost connection [18:55:28] I'm debating giving AFS training to people on our team in late Jan in NYC, and I'm wondering if I should open it up to outside people. The training would be 'AFS Admin' (e.g., would likely re-use Alf's material, with his permission, of course) for that -- for most of our engineers, that would be review material and 'filling in some gaps', not completely new material, and then 'AFS Intermediate Admin' and/or 'AFS Development Fundamentals'. Would there be any benefit in opening that training up to outside people? [18:57:23] I think so [18:57:36] I learned a helluva lot when I took IBM's training course [18:57:43] the hands on part of that course was invaluable [18:58:16] tho, I know nothing about the quality of this Alf's stuff [19:01:18] http://workshop.openafs.org/afsbpw08/afstut.html has the links.. [19:01:24] passwords are on that page. [19:02:07] it's ~175 slides of materials..not a lot different from what I've seen before (I'm somewhat familiar w/Alf's because I did an editing pass through that material for a client last yr, so it's easy enough for me to cover) [19:03:11] I'm mostly wondering: 1- is there demand for some training in NY in that timeframe and 2- would that interfere with other plans (e.g., if the Workshop was in early May in NJ again, I'd feel like it might detract from the Workshop attendance, and I wouldn't want to do that). [19:03:41] otoh, if others thought it useful, I could use the late Jan offering as a dry run to offer a more advanced class at the workshop. It just depends on what people think would be useful. [19:16:43] hi all [19:17:27] greetings [19:44:20] --- matt has left [20:27:52] it is unlikely that i'd be able to make it to nyc for training, but i would make cookies for an 'afs development fundamentals' class. [20:33:07] tx for the feedback. [20:33:48] what topics would ensure that, say, the cookies contained chocolate or were of the 'cinnamon and sugar' variety? [20:40:11] my standard bribe cookies are cliff stoll cookies with the cocoa jag option [20:40:25] with good cocoa, and good chocolate chips. [20:41:05] as for topics, i'd have to think. it's one of those topics where i whinge that such a thing existed every so often, but for various reasons. [20:41:38] the openafs codebase is fairly large and seems weirdly laid out to me. there's probably some reason to it, knowing that would be good. [20:43:48] so one topic would be a rough overview of what is where? e.g., an expansion of src/SOURCE-MAP? [20:44:33] something like that. [20:46:06] i've gotten to the point now where the layout of most major pieces makes sense to me, how a client calls go through rx, how that rx glue is generated, how the server processes mostly work, etc. that would probably be a good thing to have. [20:46:23] * stevenjenkins nods. [20:46:40] although, when it comes to the cache manager, well, i get a headache every time i try to look at that. [20:46:50] yeah, the bi-directional: "given an RPC, where is the client using it; where is the server-side piece", and working backwards from those. [20:46:54] heh. [20:47:04] jaltman's coverage of the Windows CM is quite good, IMO. [20:47:22] a bit too simplistic for developer-training purposes, but a good overview. [20:48:00] yeah, one of these days doing afsy coding things will bubble back to the top of my Project List. i really want to finish my python afs dump stuff, as well as try to make a python equivalent of the perl AFS:: modules. [20:48:30] python version of Norbert's AFS::, Alf's AFS::, or Phil Moore's AFS::Command ones? [20:48:55] Norbert's. at least, thats the one i've been using a lot at work. [20:49:12] really? [20:49:24] i think so. [20:49:37] Interesting. [20:49:38] i can never remember the family tree of those things. [20:49:46] --- dmontuori has left [20:49:51] do you: use AFS; or use AFS::Command? [20:50:02] i do not use AFS::Command. [20:50:09] that seems a bit messy to me. [20:50:13] ok, then that's norbert's. [20:50:38] the 'call out to the shell' is kludgy for Phil's, but it makes porting pretty easy (ie, trivial) [20:54:34] how important would be covering all of the various subsystems be? ie, would it be better to cover one or two in detail (and if so, which ones), or cover more breadth? [20:56:40] that probably really depends on what what the group taking the training wants out of it. i'd probably default to bredth [21:09:07] the layout is occasionally historical, and only has been reordered in fits and starts [21:09:21] with git, preserving history across reorg will be easier [21:17:27] > if the Workshop was in early May in NJ again [21:17:38] well, given that it will be at stanford june 1-5... [21:19:22] your take is..? [21:19:46] on? [21:20:52] 3 pieces: 1- Intro to AFS, 2- Intermediate AFS, 3- Intro to AFS Development. My current audience is our engineers, but I'm wondering-out-loud-and-in-public here if I should make it open to others. [21:21:42] for our guys, #1 would be really, really fast, and most of the emphasis on 2 & 3. But if others are attending, I could slow #1 down and possibly do some other bits and pieces (e.g., set up actual labs, exercises, etc). [21:22:10] there's probably value in doing so. as long as you get the details right. like, no, turning off keepalives is not a good idea. [21:22:56] I'd do my best to confuse people...after all, that would help drive more people to Alf's training. [21:23:07] oops, sorry, that wasn't supposed to be public knowledge. [21:23:09] ;) [21:23:24] i suppose as the guy who can't do simple bitshift math i should shut up [21:23:57] * stevenjenkins shrugs. I've written dozens of courses. Errors happen. [21:24:37] i've written dozens of sentences. some of them were true. [21:24:47] you do your best, get feedback and input from others beforehand, and fix problems once you're aware of them. [21:25:32] yeah, the bits I'd be most concerned about are those that recommend certain options..e.g., like what was discussed in the open discussion at the workshop this yr. [21:25:57] and, for me personally, ignorance of stuff that's newer than ~10 yrs ago. [21:27:44] btw, has max or mmeffie touched base w/you recently on the git stuff? [21:28:46] meffie gave me a pointer to their work. i then got busy making 1.4.8 happen [21:29:08] I talked w/max about it again today. [21:29:27] tried to put a sense of urgency there..and told him to sync up with you. [21:48:11] well, the ball is in my court now, probably [21:49:17] ? [21:52:30] meffie sent me blah blah and i didn't look blah blah openafs 1.4.8? [21:54:19] got it. [21:54:42] I'll ping max again tomorrow, but he's pretty busy with some DAFS joys (he was also sick virtually all of last wk) [21:55:43] i need to bug him, he was supposed to tell me what his address was, and confirm what size t-shirt his wife took [23:03:48] --- reuteras has become available [23:17:47] --- Russ has left: Disconnected [23:51:56] --- Manfred has become available