Posts by walli

1) Message boards : Number crunching : Peer certificate cannot be authenticated with given CA certificates (Message 97069)
Posted 31 May 2020 by walli
Post:
Yep, sorry that I didn't check twice, but somehow the escaping backslashs have been substituted away...

---

This forum seems to kill them for some reason, even inside a code block. Wait a sec...

It should look like:
https://pastebin.com/avKg94bv

Sorry, I cannot edit the old post to correct this :(.

All the "sed" command does is putting a "!" in front of the corresponding line.

The "!" in front of an entry marks a certificate as being deactivated, but you have to apply the changes via a "sudo update-ca-certificates". You could have put a "#" as well (which means "this is a comment"), but this would NOT remove any old existing symlink in "/etc/ssl/certs/" which is also necessary to get rid of the expired cert.
2) Message boards : Number crunching : Peer certificate cannot be authenticated with given CA certificates (Message 97048)
Posted 31 May 2020 by walli
Post:
Hi,

to all the Ubuntu guys: Playing around with "/usr/local/share/ca-certificates" is the wrong way to get this fixed. It might be possible that, depending on the order of the certs inside "/etc/ssl/certs/ca-certificates.crt" and/or the version of libcurl, you have to deactivate the expired system-wide cert to get the Boinc client running. This is done by:

# sudo sed -i 's/mozilla/AddTrust_External_Root/!mozilla/AddTrust_External_Root/' /etc/ca-certificates.conf
# sudo update-ca-certificates
Updating certificates in /etc/ssl/certs...
0 added, 1 removed; done.
Running hooks in /etc/ca-certificates/update.d...
done.

Now the old cert is not linked within "/etc/ssl/certs/" and not present inside "/etc/ssl/certs/ca-certificates.crt" any longer and the Boinc client works without any restart.

My short tests showed me that Ubuntu 20.04 and 18.04.4 work flawlessly out of the box without the above fix, Ubuntu 16.04.6 though needed this procedure. On all these 3 systems both the old "AddTrust_External_Root" and the new "COMODO_RSA_Certification_Authority" were already installed. There was no need to add the new COMODO cert (also called "1720081") to any of my systems.

Windows users please note: https://github.com/BOINC/boinc/issues/3789#issuecomment-636400593
That means: If your client is no older than 7.9.1/7.10 (so it was shipped with the latest "ca-bundle.crt" from January 2018), it's sufficient to delete the "AddTrust External Root" block in "ca-bundle.crt" and you are done. But be aware: I did not check the exact Boinc client version numbers, they are a guess based on the source code repository!

You might adapt this on other Ubuntu versions and/or Debian based distributions.

Hth and best wishes,

walli
3) Message boards : Number crunching : Aborted work and a lot of wasted time - again (Message 95302)
Posted 24 Apr 2020 by walli
Post:
My bad, apologies. :)
4) Message boards : Number crunching : Aborted work and a lot of wasted time - again (Message 95297)
Posted 24 Apr 2020 by walli
Post:
Hi guys,

We used to grant claimed credit to all canceled


Nope, see above or the other big thread where this topic is currently on discussion - people got no credits for the server-side aborted tasks... https://boinc.bakerlab.org/rosetta/forum_thread.php?id=13852

past deadline


Yes and No. I had a bunch of tasks about 2-3 weeks ago ("rb*" resp. "*robetta*"), which got no credits at all as soon as they were past the deadline even for a single second. There was no other info/hint in the stderr-log which leads to any other assumption but the deadline.

and invalid results


I'm pretty sure that this ist not the case, but I have no work units/tasks to look at atm to confirm this because the results are purged very fast.

I think this task was lost after our last server update.


Which "task"? Do you speak of a setting or something like a cronjob for this credit-problem...?

Thanks for your reply and best regards,

walli
5) Message boards : Number crunching : Aborted work and a lot of wasted time - again (Message 95210)
Posted 23 Apr 2020 by walli
Post:
Hi guys,

I just noticed that hundreds of my work units were aborted by the server yesterday, and I'm talking not only about not-yet-started ones, but tasks which already ran for hours. Altogether I lost about (2816696.11 seconds) 3.26 days of work/runtime!

I understand if you don't need the results anymore (they were *all* "aborted by project - no longer usable"), but then please go at least for some credit-compensation in the future...

Thanks,

walli






©2021 University of Washington
https://www.bakerlab.org