Hi, I just pushed version v5020 to github (https://github.com/RIPE-NCC/ripe-atlas-software-probe/tree/v5020) The main change is to the 'SOS' mechanism. Probes look up a special DNS name just before they try to connect. This can be used to diagnose probes that have trouble connecting. For hardware probes we rely on the probe's ethernet address to distinguish between probes. For software probes that doesn't work because we don't record ethernet addresses for software probes. We made a change to have software probes send the hash of the probe's public key instead. With this change the network tab of the probe page shows the 'SOS History'.
From this release we now also support CentOS 8 binary packages, though I noticed that we did not yet add that to the installation instructions.
Finally, we changed the way we accept SSH public keys. We now accepts longer RSA keys (max. 4096 bits) as well as ECDSA and ED25519 Philip Homburg
On 01 May 2020, at 14:12 , Philip Homburg <philip.homburg@ripe.net> wrote:
We made a change to have software probes send the hash of the probe's public key instead. With this change the network tab of the probe page shows the 'SOS History'.
Interesting “side effect” when using LXC containers that get their |uptime” from the running host: Resolver Time (UTC) Query Type Power-up Time Info 216.66.87.134 2020-05-03 21:59:43 AAAA 25d 0h 10m 216.66.87.134 2020-05-03 21:29:45 AAAA 0h 4m Between those I transferred (backup & restored) the container to a different host. Thus the first SOS at 21:29:45 was on the old host, that I just booted up, while the 21:59:43 was the new destination host that had been up for ~25 days :) Perhaps you could consider to include a “daemon start up timestamp” together with a “host” uptime, as it might give indications or restarted daemons too?
On 2020/05/04 0:19 , hvjunk wrote:
Perhaps you could consider to include a “daemon start up timestamp” together with a “host” uptime, as it might give indications or restarted daemons too?
Yes, it make sense to switch to the start of the Atlas service as starting point for uptime for software probes. I created a todo item to look into this. Philip
participants (2)
-
hvjunk -
Philip Homburg