Hi,
we have been using BOINC for an private project for some time. Our project is mostly aimed for Grid computing. NOw we have started to wonder if there is some way how we could turn off or configure back-off timers. From time to time we have some maintenance or just some network issue when we get some back-off on our nodes. It is a bit unfortunate. We have some fat nodes with mutiple GPUs which are then idle.

So to my question. Is there some way how to adjust the backoff increments on BOINC client or server or ideally is there some way how to disable this feature or at least set it up to some constant time? (I've checked the cc_config and boinccmd documentation but I might be missing smth)

Thanks in advance.

For the server you can set next_rpc_delay in config.xml, and you ought to be able to increase that prior to the server going into maintenance, to an amount of time the server is off line. Time is in seconds.

The client's internal backoff timers cannot be overridden. Unless you go through the source code and take them out, then rebuilt. The timers have a random value with a maximum of 24 hours. Although you can use "Retry pending transfers" in BOINC Manager to temporarily try to make contact. For boinccmd that's --network_available (and should be scriptable).

---

There is also

<maintenance_delay>nseconds</maintenance_delay>
If the project is down, tell clients to delay their next request for at least the given number of seconds.

If you often have short outages (below the default 3600 seconds, or 1 hour), that would allow work to restart quicker.

But be aware of the downside: if you have a longer outage, that will allow the clients to make many more failed attempts to contact the server, and the random client backoff mentioned by Jord will be more likely to reach the maximum 4 hours.

Bookmark Project configuration for a long list of tuning options.

Thank you for a quick response. We will try to fix our issue with the tips you have provided.