Looking at our broker metrics, I don't see anything obviously wrong here. It looks like our pool of unrestricted proxies is being overwhelmed by client polls. Although it does look like our unrestricted capacity did drop around the 25th (when we were originally having issues with the NAT probe service and it hasn't recovered in the same way that the restricted capacity did.

One reason for this might be that standalone snowflake proxies only perform the NAT check on startup and not every 24 hours like the web-based proxies do. So instead of recovering within 24 hours, they will wait until restart to recover. I created an issue for this: https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowf...

We might also be having issues with the NAT probe service failing to provide correct or definitive answers for unrestricted proxies. I'll defer to the existing issue for work on this: https://gitlab.torproject.org/tpo/anti-censorship/pluggable-transports/snowf...

Cecylia

On 2021-11-02 10:42 a.m., alertmanager@hetzner-nbg1-02.torproject.org wrote:

Total firing alerts: 1 Total resolved alerts: 0

## Firing Alerts

---

Time: 2021-11-02 14:41:30.286335746 +0000 UTC Summary: There are too many clients not finding a proxy to connect Description: For each connected client 1.7 couldn't find a proxy to connect

---

##Resolved Alerts

---

anti-censorship-alerts mailing list anti-censorship-alerts@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/anti-censorship-alerts