Old scheduler URL is taken offline
25 Jul 2017 13:27:18 UTC
Topic 208929
(moderation:
Update 08/02/2017 14:00 UTC: Using the old URL now causes a "404 not found" error send to the client which should then download the new scheduler URL, see below.
Hi,
some of your computers are still using an old scheduler URL to communicate with Einstein@Home. We are soon disabling the redirect that was put in place to transition to the new scheduler URL. Unfortunately there is no easy way to find out if one of your hosts is affected.
NOTE: This is not the URL you use to add Einstein@Home as a project! The scheduler URL is different from the project URL and the latter does not need changing.
This transition will happen automatically but will take some time during which your computer is unable to report finished work or get new work. Here are some possibilities on what you can do to speed this up:
- Do nothing. If your host is affected, it will download the new scheduler URL after 10 failed attempts to contact the old scheduler URL which may take up to 2 days (when running 24/7) because of the exponential backoff mechanism.
- Click the "Update" button in BOINC Manager multiple times to force failed project communications. After 10 failed attempts the Client will automatically download the new URL (visible in Manager Event Log, look out for master file download from Einstein@Home).
- ADVANCED users only: Replace the old URL with the new in client_state.xml. Shutdown the Client and open client_state.xml in an Editor. Look for "<scheduler_url>http://einstein5.aei.uni-hannover.de/EinsteinAtHome_cgi/cgi</scheduler_url>" and replace with "<scheduler_url>https://scheduler.einsteinathome.org/EinsteinAtHome_cgi/cgi</scheduler_url>". Save the modified file and start the Client again.
A lot of my hosts have been
)
A lot of my hosts have been running for quite a long time and although there were quite a few that had hardware upgrades (mainly graphics cards) I'm still using the same OS version and BOINC version on most of them. There have been eight where the GPU was an AMD RX 460 which required a brand new OS install and the OpenCL components from the AMDGPU-pro driver in order to crunch the current tasks. My gut feeling was that the majority of my hosts might be using the old scheduler URL.
In order to simplify the process of forcing a change when the redirect is removed, I've written a module to do it automatically without having to stop BOINC. I had a 'utilities' script that already existed for doing housekeeping on the fleet (things like deleting old cached data files, tweaking max and min work cache settings, making CPU/GPU run mode and network mode changes, etc). The script is constructed in such a way as to allow new functionality to be added (in the form of an additional module or procedure) quite easily.
The new module does two things. It produces a list of hosts that are still using the old scheduler URL. Then it can use that list and cycle through it using boinccmd to force a series of 'updates' on each host until a "master file download" is detected in stdoutdae.txt. It repeats this on each successive host in the list. I've just run the new module to create the list with the second stage being disabled.
I have 82 active hosts currently under script control. The new module found only 31 having the old URL and 51 having the new. I was quite surprised at how many had already made the transition.
Cheers,
Gary.
I just deactivated the old
)
I just deactivated the old URL. Please see my first post on how to proceed. Please post a log snippet you find problems with my advise.
I've run my script and all
)
I've run my script and all hosts now have the new scheduler URL. No problems were noticed. Once the new master file download occurred, all was well. Some hosts had quite a few tasks to report and quite a few new ones to receive. There didn't seem to be any issues.
Thanks for the early heads-up about this.
Cheers,
Gary.
Thanks for checking Gary and
)
Thanks for checking Gary and reassuring to hear that it works as I expected.