[07:11:00] --- reuteras has left [08:45:24] --- Claudio Bisegni has become available [08:45:29] --- Claudio Bisegni has left [10:12:07] --- tkeiser@sinenomine.net/owl has become available [13:13:00] --- haba has become available [13:13:57] --- mansaxel has become available [13:14:14] good evening [13:14:32] (this is europe, where evening is now) [13:14:53] mansaxel and I are wondering how to best fake "this is a good /vicepa (inode)" for a salvager that likes to core. [13:15:24] We just want the volserver to get up to come to a vos dumpable state. [13:15:58] we've got an old and barely resurrected fileserver that needs killing, but would want to grab them volumes. [13:16:10] (no, no backups...) [13:28:13] interesting question. have you tried just running a volserver process on it, no fileserver? [13:28:45] nope. Not tried that. Will do... [13:29:33] I've never tried that, btw. [13:29:42] so I have no idea if it would work or not, but it's a thought. [13:36:15] looks like it wants to talk to something on startup, my guess is a fileserver. [13:41:56] a DAFS volserver will need a fileserver to act as syncronization point, but if you're running inode, that implies non-DAFS. [13:42:19] I don't recall off-hand what a 1.4 volserver does, though. [13:42:48] This is 1.4.0 unfortunately [13:48:02] ah, it is trying the FSYNC connection to the server. [13:49:12] 1.4.x also has (a greatly simplified) fssync [13:49:21] have you tried vol-dump? [13:49:31] gotta go. be back in ~1hr [13:54:53] haba, it would be helpful to know why the volserver is failing. what does the core tell you? [13:55:07] just a moment [13:56:30] to me, one interesting topic for an afs project (summer of code?) would be storing metadata about volumes, folder, and files [13:56:56] what kind of metadata and where? [13:57:09] and how do you want to access it and for what? [13:57:16] if I'm building vol-dump, is it better to build latest or the release I'm trying to extract from (1.4.0?) [13:57:50] --- Russ has become available [13:57:57] there have been no data format changes in 1.4.x [13:58:04] ok [13:58:14] *grabs latest* [13:58:16] 1.4.8 fixes bug that were present in 1.4.0. [13:58:47] if the latest binaries core we would like to fix it [13:58:54] ok [13:59:03] secureendpoints - some of that would have to be fleshed out more... but it's a direction that research communities as a whole are going [14:01:08] I'm trying to understand, do you want metadata about the afs namespace itself? in which case the tools that JPL developed for managing export control issues are what you are looking for. Or perhaps you want extended attribute and multiple data stream support in AFS so that you can use AFS to store arbitrary application specific metadata. [14:01:39] if its the former, I'm trying to get those tools open sourced. [14:01:45] Anyone got a working linker line for vol-dump [14:01:48] if it is the latter, it requires protocol changes [14:01:52] I'm not exactly sure what you're referring to with the former [14:02:04] I"m thinking more along the lines of the latter [14:02:29] say I"m looking for files from years ago related to "Project X" [14:02:35] (the Makefile gives me gcc -pipe vol-dump.o -o vol-dump which is not quite enough) [14:02:40] so, I do something like "fs showmeta 'project x'" [14:02:48] and it shows me all those files with that metatag [14:03:00] taht's a simple example, tho [14:04:44] to perform a search against an arbitrarily large name space requires a search engine and in the case of an access controlled name space would require indexes based upon the privileges of the requestor. [14:04:51] I understand what you are looking for. [14:05:34] yes, exactly [14:05:52] this is an interesting problem, and many different institutions are trying to solve this problem [14:05:58] differently. [14:06:04] You want to be able to tag arbitrary data at the file, directory or volume and have the data be collected in a database independent of the file system such that a query can be issued to the database to return a list of objects that match the tags [14:06:33] but usually what folk come up w. are these clunky web-based things that are good for searching but not for actually working with files [14:06:49] yeah [14:06:52] I gave a talk on this subject to Nokia a few months ago. [14:06:55] a la spotlight for afs [14:07:20] we as Cornell and as research units at cornell are starting to look at these problems [14:07:24] I would start by implementing Apple double file support in the clients [14:08:40] well, one does not want to have to search through the whole filespace [14:09:14] tho, spotlight is not necessarily the right answer... but its indexing some of the contents of files is also interesting [14:09:19] but the long term solution requires modifying the data structures used to store file, directory and volume data so EAs and streams can be natively supported in the FS-CM protocols. The file server would then need to be backed with a database to perform the indexing. [14:09:35] yeah... a db definitely has to be involved somehow [14:10:21] if the metadata is collected by the file server and summarized for each volume. the searching can be quickly performed. [14:10:29] yep, exactly [14:10:51] I think such a project can be prototyped without touching the existing afs protocols. [14:11:11] there are a lot of open questions in this space [14:11:18] also, it might be a good GSOC project, imho [14:11:24] there are [14:11:39] that would need to be answered before protocol and data format changes for a production file system are proposed [14:11:49] oh, btw, would gcc suffice for building userland stuff on Sol10? [14:11:56] do you want to mentor this project? [14:12:01] I'm happy to, yes [14:12:27] good, then we will take this offline and work up a project proposal. The deadline is 3/9 [14:12:47] ok... just lemme know what I need to do [14:14:22] start off by writing up a three paragraph description of what you would like to have achieved over the Summer as part of the project. [14:15:20] short or long paragraphs? [14:15:23] you are most likely getting 10 weeks, 30 hours a week from a sophomore with little real world development experience [14:16:34] try scope the project such that the success criteria can be flexible depending on how quickly the student picks up new things. [14:17:05] I would say as long as the paragraphs need to be to express the idea. [14:17:19] k [14:18:40] when you are done send it to openafs-gatekeepers@openafs.org and we will work with you to refine the proposal and recommend some approaches that might work out best. [14:18:56] ok [14:21:00] please be aware that mentoring is a non-trivial amount of work and volunteering should not be taken lightly. last Summer the mentors spent somewhere between 10 and 15 hours a week working with their students [14:21:09] ok [14:21:36] and all you get in return is a lousy T-shirt and if you are really nice some free beer in October [14:22:22] I got the following line under Linux (we'll see what happens when mansaxel gets there on Sol10): gcc -pipe -o vol-dump vol-dump.o ../dir/libdir.a physio.o libvolser.a ../../lib/libaudit.a ../../lib/vlib.a ../../lib/libacl.a ../../lib/libsys.a ../../lib/libvldb.a ../../lib/libubik.a ../../lib/libauth.a ../../lib/libcmd.a ../../lib/librxkad.a ../../lib/libdes.a ../../lib/librxstat.a ../../lib/librx.a ../../lib/liblwp.a ../../lib/libsys.a ../../lib/libcom_err.a ../../lib/libkauth.a ../../lib/libusd.a ../../lib/util.a -lresolv [14:24:58] the link line from your volserver should be sufficient. [14:25:19] eg, cut and paste from the Makefiles [14:26:40] I want a SunStudio on that box anyway. [14:27:10] The above line was sufficient. The one I did oversee at the first try was physio.o. [14:27:58] s/gcc/cc/ and s/-lresolv/whatever-sol10-needs/ [14:57:10] Ok. I'll sleep. Thanks for all the help. [15:08:48] --- haba has left [16:06:05] grrr... there's no dinner waiting here for me [16:08:03] is there normally dinner waiting in the openafs conference room? [16:08:13] no [16:08:17] but I came home hungry [16:08:23] now I have to figure out wtf to do for dinner [16:08:29] maybe we can all chip in and hire a chef [16:09:59] *looks in fridge* [16:16:27] *finds grillers* [17:13:09] --- Russ has left: Disconnected [17:15:45] --- edgester has become available [17:43:53] --- edgester has left [23:17:44] --- reuteras has become available