Hello,

DepSpid is using DynDNS and so it gets everyday a new IP adress, afaik. Apparently BOINC can not handle this and you have to stop and restart it before it reports finished results and downloads new work (after the IP of the project changed).

My question: Is there any way to update DNS / IPs for BOINC without restarting it? Are those IPs stored at a place which is accessible?

I've tried a 'flushdns' which didn't help and I searched through some of the xmls in the BOINC directory but did not find anything obvious.

If you run BOINC only on one computer and all other projects are checkpointing then it is not a big issue to restart BOINC once a day. But if you have a number of PCs (as I do, and some of those have no monitor, keyboard or other needless stuff) then it can be quite annoying.

Any solution (including not-so-kosher hacks)? Thanks in advance.

Regards

Alex

Excerpt from the messages:

18.01.2007 20:00:00|DepSpid|Sending scheduler request: To fetch work
18.01.2007 20:00:00|DepSpid|Requesting 1 seconds of new work
18.01.2007 20:00:22||Project communication failed: attempting access to reference site
18.01.2007 20:00:24||Access to reference site succeeded - project servers may be temporarily down.
18.01.2007 20:00:26|DepSpid|Scheduler request failed: couldn't connect to server
18.01.2007 20:00:26|DepSpid|Deferring communication 2 hours, 30 minutes and 57 seconds, because scheduler request failed
18.01.2007 20:34:25|DepSpid|[file_xfer] Started upload of file spider_24215_0.xml
18.01.2007 20:34:48||Project communication failed: attempting access to reference site
18.01.2007 20:34:48|DepSpid|[file_xfer] Temporarily failed upload of spider_24215_0.xml: system connect
18.01.2007 20:34:48|DepSpid|Backing off 47 minutes and 24 seconds on upload of file spider_24215_0.xml
18.01.2007 20:34:50||Access to reference site succeeded - project servers may be temporarily down.

---

I keep trying to attend a meeting of quantum physicists, but every time they agree on a location the appointed time changes and vice versa.


SETI.Germany: 0.3 billion credits

I can tell you that this issue has been discussed on the mailing list, but I don't think there is an immediate solution. I didn't pay very close attention to the thread.

Anyone else???

There are reports of success and failure with the new alpha client which has the new version of libCurl in it.

I'll look more closely at it today. I've got a couple units waiting to upload and the server is offline right now. So I'll see if they upload or if I have to kick start BOINC with a file-exit to make it work.



This thread from the project board is where most of the discussion is going on. Third page has a post with a way to check what IP address is being contacted by BOINC and what IP address the server is actually using.

---

Kathryn :o)

Thanks a lot for your replies, Kathryn and Aurora B.

I read the thread over at the DepSpid forum and did the suggested tests with TCPview and nslookup. I tested it with 5.4.11 on one system and 5.8.1 on another (which is what I had installed up to now). It's quite obvious that BOINC is using an 'old' IP instead of the current one (today 82.83.207.33) until it is restarted.

Now I have upgraded the 5.8.1 installation to 5.8.3 (which does indeed use a new libcurl: 7.16.0 instead of 7.15.5). Let's see what happens tomorrow.

Furthermore, I played a little bit with the client_state files, namely the tag

<scheduler_url>http://depspid.dyndns.org/cgi/index.cgi</scheduler_url>

there. I suspended cpu and network activity in BOINC (without closing it) so that it doesn't update the client_state files anymore. Then I changed "depspid.dyndns.org" to the current IP adress and cancelled suspension again. BOINC updated the client_state files and wrote the original domain name into it, overwriting the changes.

So, I guess, BOINC is fetching the scheduler url from the master_url on each connection attempt, right?

If yes, another idea, if you don't mind:

The homepage of DepSpid currently has the following scheduler tag:

<!--
      <scheduler>http://depspid.dyndns.org/cgi/index.cgi</scheduler>
-->

What would happen, if a second tag would be added which uses the current IP adress instead of the domain name? Something like this:

<!--
      <scheduler>http://depspid.dyndns.org/cgi/index.cgi</scheduler>
      <scheduler>http://82.83.xxx.xxx/cgi/index.cgi</scheduler>
-->

If I understand this page from the BOINC website correctly (see the paragraph titled "The master URL"), then it is possible to have more than one scheduler url and BOINC would parse them all. Would BOINC even accept an IP address instead of a domain name? Most TCP/IP programs do, but you never know... ;-)

I am aware, that the IP address would have to be changed every day, but if this can be done by a script, it might not be a big deal.

Am I completely off or is it worth a suggestion?

Regards

Alex