My models crashed when my computer lost power. I had a backup of the BOINC directory, so I just deleted the old one and overwrote it with the backup. I got a message that the computer generated a new cross project id, and now the tasks are labeled as Client detached on the website. They are still running on the computer.

Did I do something wrong?

No, not at all. What happened is your client delivered trickles to the server after the backup was made, and the BOINC system cries foul when machines appear to time travel. This is not a problem for CPDN, and you'll still get full credit for the models if you finsh them. To the researchers they're the same as any other completed model.

The only way to prevent this from happening is to restore from a backup made before more trickles reach the server. Something like limiting communication to a specific hour every other day with a backup immediately following.

Each time the computer's BOINC client contacts the server, a counter is incremented to record the number of the contact. This counter is <rpc_seqno>. (One for each project.)

A way around the problem with backups, is to make a note of the <rpc_seqno> for the cpdn project BEFORE you delete the original set of files. (Or save a separate copy of client_state.xml so that you can look at it after deleting.) This is under the <project> section of client_state.xml.

Once you've restored the backup, and BEFORE re-starting BOINC, change the <rpc_seqno> to what was in the original file. Then the server won't know anything about the files being changed.

---

Backups: Here

That sounds good Les. I'll give that a try next time I restore from backup.

Another quick question. What if you have to restore a backup that is several days old, so as the model runs it will try to resubmit trickles that have been already submitted. Is there any problem with that?

No. Duplicate received trickles are ignored.

---

Backups: Here