- noctool-devel@lists.common-lisp.net - mailman3.common-lisp.net

[noctool-devel] what is a cpu-monitor?
by Jim Prewett 16 Jun '08

16 Jun '08

Hi Ingvar, What is the CPU monitor supposed to do? I can't find a PROCESS method for it. :P :) I'd be happy to try to write such a method if only I could grok what you're thinking here. :) Jim James E. Prewett Jim(a)Prewett.org download(a)hpc.unm.edu Systems Team Leader LoGS: http://www.hpc.unm.edu/~download/LoGS/ Designated Security Officer OpenPGP key: pub 1024D/31816D93 HPC Systems Engineer III UNM HPC 505.277.8210

2 4

[noctool-devel] More network protocol things
by Ingvar 15 Jun '08

15 Jun '08

Some thinking later and I've split the network code into three files. If anyone wants to play around with it, I have a (temporary) extra ASDF definition to make it possibel to load noctool and noctool-with-network separately (makes it slightly easier to find network-specific compilation issues), I'll paste it in towards the end. Now, what I've done, to make the protocol dispatch easier (hahaha) to follow, is a dispatch table, keyed off the protocol message header and a macro taht defines handlers. The defhandler macro looks like this: (defhandler <function name> (<protocol header> <rest> <of> <message>...) ... function body goes here) This does two things, it defines a function <function name>, with CONN as the first argument (in most cases, you need to know what connection to send a reply back to) and the message type thingie declared as ignorable (since, most of the time, it's only used for dispatching, anyway). Not very tested, as of yet, I'll probably need to frob around with how to retrieve and set proxy objects, but... And here's the system definition (once the network code is a bit more stable and "with it", I'll merge this into the main system definition, I won't check this temp definition into CVS): (in-package :asdf) (defsystem "noctool-network" :depends-on ("noctool") :components ((:file "network-globals") (:file "network-utils" :depends-on ("network-globals")) (:file "network" :depends-on ("network-utils" "network-globals")))) //Ingvar

1 0

[noctool-devel] We (almost) have network!
by Ingvar 14 Jun '08

14 Jun '08

I've successfully managed to connect one noctool image to another noctool image and identify one way. Next step, make sure that everything seems to work and make it happen both ways. Once that is in place, it's time to start testing proxy objects and the like. After that, it's time to modify the scheduler slightly (well, the PROCESS method(s)), to send updates across the wire on an as-needed basis. //Ingvar

1 0

[noctool-devel] image.asd fixes
by Jim Prewett 12 Jun '08

12 Jun '08

Hi Ingvar, While using ASDF-INSTALL to get a new noctool setup going, I found that the image package used strings instead of symbols for the packages that it depends on. This confuses the heck out of asdf-install (but seems to work OK for ASDF - HMMM... sounds like it might be a bug in ASDF-INSTALL, but... well, I don't want to spend the time reading the spec to determine that ;P :) I've attached a modified version of image.asd from image-0.6 - the changes are by no means profound :) Using the attached image.asd made my asdf-install go very smoothly. :) hth, Jim James E. Prewett Jim(a)Prewett.org download(a)hpc.unm.edu Systems Team Leader LoGS: http://www.hpc.unm.edu/~download/LoGS/ Designated Security Officer OpenPGP key: pub 1024D/31816D93 HPC Systems Engineer III UNM HPC 505.277.8210

2 3

[noctool-devel] anonymous CVS works!
by Jim Prewett 12 Jun '08

12 Jun '08

Hey All, I found that anonymous CVS access (using the directions provided on the main noctool webpage) works! Wha-Hoo! Ingvar, do you want to modify that webpage to remove the statement that it is not yet working? Jim James E. Prewett Jim(a)Prewett.org download(a)hpc.unm.edu Systems Team Leader LoGS: http://www.hpc.unm.edu/~download/LoGS/ Designated Security Officer OpenPGP key: pub 1024D/31816D93 HPC Systems Engineer III UNM HPC 505.277.8210

2 3

[noctool-devel] compact configurations for identical machines
by Jim Prewett 05 Jun '08

05 Jun '08

Hello, I was thinking about what a bear it will be to have to specify each of my compute node systems in my clusters individually. I was thinking about syntax such as the following that would create an equipment object for each of 256 hosts named l001 through l256. (machine-cluster "LosLobos" linux-host (user "download") (disks (disk "/dev/sda1" 80 95)) (machinerange "l001-l256")) What do y'all think? Is there a better way to do this? Something like this would reduce the size of my shop's config file a couple orders of magnitude - in turn, I'd screw things up accordingly less often :) Jim James E. Prewett Jim(a)Prewett.org download(a)hpc.unm.edu Systems Team Leader LoGS: http://www.hpc.unm.edu/~download/LoGS/ Designated Security Officer OpenPGP key: pub 1024D/31816D93 HPC Systems Engineer III UNM HPC 505.277.8210

3 20

[noctool-devel] My current work
by Ingvar 05 Jun '08

05 Jun '08

I thought I'd knock the network bits into order and get some testing done. There's been some check-in thanks to that (mostly in config.lisp and packages.lisp, so there's a way of configuring remote nodes). Once that's done (unless someone preempts me), I'll write a wrapper to run NAGIOS call-outs. Can't say I remember off-hand how they interface, but I wrote some a couple of years ago, so a bit of searching should take care of that. If anyone wants to preempt me and do that I won't complain. Oh, with "done" I mean "can successfully make a network connection and authenticate", not "all of the network protocol is implemented". :) //Ingvar

1 0

[noctool-devel] "How would you like to have your ad on 2 Million Websites ?"
by BlogBlaster＠common-lisp.net 25 May '08

25 May '08

How would you like 2 Million Sites linking to your ad ? Weblog or blog population is exploding around the world, resembling the growth of e-mail users in the 1990s. Post your ads where people read them! - What if you could place your ad on all these sites ? Right, that would mean you would have millions of sites linking to your ad. For Full details please read the attached .html file Unsubscribe please read the attached .html file

2 1

[noctool-devel] SNMP support?
by Ingvar 19 May '08

19 May '08

Icky as it is, I fear we'll be needing to tackle the horribleness that is "SNMP support", sooner or later. Makes it super-icky supporting routers and switches, otherwise (yes, to an extent, you can extract SOME diagnostic data via command lines, on quite a few different routers and switches, it's just not necessarily trivial). This is, FWIW, where things ground to a halt last year (SNMP support and my not having multiple boxes, so I could actually trial the network protocol in a sensible way). This is what I want/need from a suitable SNMP package: + ASDF-INSTALLable (I don't think we can compromise here) + Ideally written completely in Lisp (that is just a preference, not something that is needed, if the best option uses FFI, the best option uses FFI) + The results from queries are lisp-friendly (so we can process them in an assortment of ways) + Introspectable MIB trees (nice, though maybe not necessary) So, the options that I can find: + Lisp-SNMP Pros: Listed on CLiki Cons: not ASDF-INSTALLable (doesn't, in fact, even have an ASDF system definition) + cl-net-snmp Pros: hosted on common-lisp.net Cons: not ASDF-INSTALLable; depends on at least one package that is hard to find (as in "didn't work when I tried) That's about it, as far as I can see. Anyone know any other SNMP implementation? //Ingvar

1 0

[noctool-devel] Monitor-configuration question
by Ingvar 18 May '08

18 May '08

So, there's this sexp-based configuration language. And I was wondering, if we have a monitor that has lots of assorted defaults, but it's good to have those tunable, should we use keyword-like syntax? At the moment, I think the following would be good for tuning ping monitoring: (machine "blah" :linux-host (ping :max-rtt 500 :max-fail 17 :ping-count 42 :interval 3600)) That instantiates a ping monitor that should re-schedule every hour, do 42 pings, consider 17 failed to be enough for a red alert and has an RTT threshold of 500 ms. Defaults, at the moment, are "1 failed", "20 ms RTT", "60s interval" and "5 pings". Disk monitors, however, use a blend of required and optional arguments, but it makes sense there, because without specifying anything at all, there will be a set of default disk monitors and the main reason you'd want to configure anything is to change the "percentage used space" threshold. But, then, this makes the configuration language somewhat irregular. Though I personally prefer "makes sense" config language to "super-regular". Thoughts, anyone? We're not at a point where we have to commit 100% to the config language, but it'd be good to have something more-or-less nailed down the day we build the first release. //Ingvar

1 0