Posts by mgruben

1) Message boards : Number crunching : How do I get my Weak Key (Message 76756)
Posted 22 May 2014 by mgruben
Post:
Title says it all!

Where do I find my Weak Key for Rosetta@home? Seems in 2011, there was a mention of upgrading the server software to support them.

So how do I find it?
I spent some time a few months ago looking for the weak key without success; if you're looking for it for the purpose of attaching a computer to your account via the command-line interface, I suggest you just follow the
boinccmd --lookup_account URL email passwd
boinccmd --project_attach URL auth
path, as unsatisfyingly less-secure as that is.
2) Message boards : Rosetta@home Science : feature request - linux viewer - preferably java based (Message 76688)
Posted 3 May 2014 by mgruben
Post:
today rosetta linux participants could only basically run the jobs/tasks with little clue as to what is going on. unlike windows participants who benefits from a cool gui, it'd be great if rosetta@home project could develop a gui say based on java so that more platforms could benefit with a gui of this state-of-the-art project :)

Not officially supported, but check out boinctui if you'd like a command-line interface based task viewer
3) Message boards : Rosetta@home Science : designing novel, super-stable scaffolds for the rapid development of protein-based drugs (Message 76683)
Posted 2 May 2014 by mgruben
Post:
Very cool, many thanks for keeping us updated on the meaning behind the work units
4) Message boards : Number crunching : Certificate Errors (Message 76620)
Posted 14 Apr 2014 by mgruben
Post:
[PotentiallyUselessSpeculation]
Might this be a ripple of the HeartBleed pebble last week? [/PotentiallyUselessSpeculation]
5) Message boards : Number crunching : Can't get tasks! (Message 76601)
Posted 9 Apr 2014 by mgruben
Post:
Ditto that, although honestly the rig from which this log was taken can't seem to fill its work-unit cache, while another rig of mine has a full cache
3424: 08-Apr-2014 20:07:11 (low) [rosetta@home] Sending scheduler request: To fetch work.
3425: 08-Apr-2014 20:07:11 (low) [rosetta@home] Requesting new tasks for CPU
3426: 08-Apr-2014 20:07:15 (low) [rosetta@home] Scheduler request completed: got 0 new tasks
3427: 08-Apr-2014 20:07:15 (low) [rosetta@home] No work sent
6) Message boards : Rosetta@home Science : Rosetta@home Research Updates (Message 76574)
Posted 31 Mar 2014 by mgruben
Post:
Ratika,

Updates are always exceedingly welcome, as I enjoy perusing the fruits of my (but mostly other people's) labor.

Would it be possible to release some kind of weekly or monthly email containing all the updates posted in that period of time? I could subscribe to this forum thread, but admittedly fora are places for multi-directional conversation, and at times I just want to hear from one speaker.
7) Message boards : Number crunching : WU Download Scheduling (Message 76493)
Posted 24 Feb 2014 by mgruben
Post:
Wouldn't you know, the one setting I hadn't tried was just what I was looking for.

That is, setting
Computer is connected to the Internet about every [2.0] days
Maintain enough work for an additional [0.0] days

Sends scheduler requests at (well, really near) the beginning of the network availability window.
It's exactly the setup I was looking for, and was the one I hadn't yet tried.
Also, this setup renders obsolete the systemd setup mentioned below, which I turned off (disable; stop)
8) Message boards : Number crunching : WU Download Scheduling (Message 76483)
Posted 21 Feb 2014 by mgruben
Post:
Sounds like a good question for the BOINC message boards.
Agreed
When you get work loaded, does it process in something pretty close to the initially estimated time? If you recently changed your Rosetta runtime preference that might throw off the BOINC Manager's estimates and cause it not to realize it is going to need work.
BOINC's estimates of time-to-completion seem to be fairly accurate, and I have been fiddling with the 'Computer is connected to the internet every … days' and 'Maintain enough work for an additional … days' options since this thread began, but consistently BOINC doesn't initiate a scheduler request until 4:30a (hence my systemd script which functionally solves my wait-until-too-late-in-the-network-availability-window problem).
9) Message boards : Number crunching : WU Download Scheduling (Message 76466)
Posted 20 Feb 2014 by mgruben
Post:
I'd suggest adding another day to your work buffer (which is done in the network preferences).

My initial qualm was admittedly one of efficiency aesthetics (that, within a sufficiently-long connectivity window, new work isn't fetched until there is insufficient time remaining to complete its download).

While your suggestion would probably accomplish the same ends (namely, having enough work on the rig so that it never idles), it differs with my update philosophy, and still leaves me with the question of how the client decides when to initiate a project update (which is consistently 30.0 minutes prior to the close of the network window; an odd artifact suggesting to me that this project-update behavior is changeable).

Edit:
For anyone hoping to use a similar setup, I actually don't recommend initiating a project update request at the absolute beginning of the network connectivity window; some projects may throw you a "Not requesting tasks: too many uploads in progress" message back. Better to wait until you think all (or most) of your uploads will finish.
Accordingly, I've changed "03:01:00" in boinc-scheduler-update.timer below to "04:00:00", which gives me enough time within my 3:00am to 5:00am network window for the majority of my uploads to finish.
10) Message boards : Number crunching : WU Download Scheduling (Message 76450)
Posted 19 Feb 2014 by mgruben
Post:
Until I find the boinc-specific setting I'm looking for, I just set up a systemd job (running Arch linux; we shun cron):

/etc/systemd/system/boinc-scheduler-update.service
[Unit]
Description=For Rosetta@Home, forces a scheduler request.

[Service]
Type=oneshot
ExecStart=/usr/bin/boinccmd --passwd [from /var/lib/boinc/gui_auth_rpc.cfg] --project http://boinc.bakerlab.org/rosetta/ update

/etc/systemd/system/boinc-scheduler-update.timer
[Unit]
Description=the timer for daily BOINC project updates (currently pegged to the start of the network availability window at slightly after 3am)

[Timer]
OnCalendar=*-*-* 03:01:00
Unit=boinc-scheduler-update.service

[Install]
WantedBy=basic.target

Then of course "sudo systemctl enable boinc-scheduler-update.timer && sudo systemctl start boinc-scheduler-update.timer"
11) Message boards : Number crunching : WU Download Scheduling (Message 76443)
Posted 18 Feb 2014 by mgruben
Post:
Agreed, but that is more of a BOINC Manager question than Rosetta.
Conceded.
If it gets started on downloads and the network window ends, then it should pick up where it left off the next night (assuming you allow the same time window each night??)
It does, but it only finishes night 1's leftover downloads at the start of night 2. It then waits until T-0:30 to request night 2's downloads, which it can't finish during night 2 and which it finishes at the start of night 3, ad infinitum.
12) Message boards : Number crunching : WU Download Scheduling (Message 76441)
Posted 17 Feb 2014 by mgruben
Post:
From the wiki,
for url in $(boinccmd --get_project_status | sed -n 's/s*master URL: //p'); do
  boinccmd --project ${url} update;
done

It would be nice if it didn't wait until it had 30 minutes left, but the above at least seems to be a quick fix
13) Message boards : Number crunching : WU Download Scheduling (Message 76439)
Posted 17 Feb 2014 by mgruben
Post:
If you are running out of work, with a small network window like that, I'd suggest configuring BOINC Manager to keep more days of work on hand setting a larger value in the additional work buffer.


I've currently set #days to 1.0 and cached (additional) to 2.0, and when it can finish the downloads, that's more than enough work for a single day; my problem is that within the 3-5a download window, it wais until 4:30 to download, then 5a comes before the downloads finish.

I used to have the window set at 3-6a, and they would wait until 5:30a to start downloading, so it seems like a 30-minutes-before-window-closed rule is coded somewhere in a config file
14) Message boards : Number crunching : WU Download Scheduling (Message 76434)
Posted 16 Feb 2014 by mgruben
Post:
Hey all, I've currently got my 2 laptops set to connect to the network only between 3a and 5a, but they pretty much always wait until 4:30a to request work for the whole day, which for them takes longer than 30 minutes to download.

Is there any way I can tell them to request work earlier in the day so they don't run out of network time?






©2024 University of Washington
https://www.bakerlab.org