Hi there,

ROS 6.43 added *) watchdog - added "ping-timeout" setting;

But this is undocumented. Please add documentation for this new setting explaining exactly how it works.

As i understand it goes with Ping Start After Boot, so when Watchdog happen it will be ping time out for 60s and then process will start over

As i understand it goes with Ping Start After Boot, so when Watchdog happen it will be ping time out for 60s and then process will start over

From what i know there's more to it. It doesn't restart after 1 failed ping, but requires a certain amount of failed pings within a time interval. This should be properly documented in the documentation and is not currently documented. See https://wiki.mikrotik.com/wiki/Manual:System/Watchdog

did you make a test? what your result say?

did you make a test? what your result say?

It would not be correct for me to speculate about how a feature works. It should be documented.

Any news on this? The feature remains undocumented, ten months after introduction.

yes please update watchdog manual.

Somehow we have missed to update wiki Watchdog page, now this feature is documented.
https://wiki.mikrotik.com/wiki/Manual:System/Watchdog

Somehow we have missed to update wiki Watchdog page, now this feature is documented.
https://wiki.mikrotik.com/wiki/Manual:System/Watchdog

Thank you for the update.

watch-address (IP; Default: none) The system will reboot, in case 6 sequential pings to the given IP address will fail. If set to none this feature is disabled. By default router will reboot every 6 minutes if watch-address is set and not reachable.
no-ping-delay (time; Default: 5m) Specifies how long will it wait before trying to reach the watch-address.
ping-timeout (time; Default: 60s) Specifies the time interval in which the device will be pinged 6 times (after "no-ping-delay").

It's still a little bit clunky.

watch-address says by default router will reboot every 6 minutes. Which makes sense if 6 pings of 60s each time out. But ping-timeout says "Specifies the time interval in which the device will be pinged 6 times" which suggests that it will send out 6 pings for every ping-timeout (default 60s) so 36 pings every 6 minutes. Is that really suppose to be 6 times for ping-timeout?

Now another question: Does the 6 minute timeout reset if any one ping succeeds? Or does it keep adding up the missed pings and then even if they are 5 hours apart when it hits the 6 timeout it restarts?

I think it is like this: after boot, the router first waits for no-ping-delay before doing anything. To give other devices time to initialize and make a connection.
Then, it starts pinging at an interval of ping-timeout/6. So, one ping every 10 seconds for the default ping-timeout of 60s.
When no pings get returned for the ping-timeout interval (i.e. 6 consecutive missing ping replies) the router reboots.

I think it is like this: after boot, the router first waits for no-ping-delay before doing anything. To give other devices time to initialize and make a connection.
Then, it starts pinging at an interval of ping-timeout/6. So, one ping every 10 seconds for the default ping-timeout of 60s.
When no pings get returned for the ping-timeout interval (i.e. 6 consecutive missing ping replies) the router reboots.

That seems quite plausible. But I really urge Mikrotik to make it absolutely clear in the documentation as this is a critical feature for routers and if this feature misbehaves or is misunderstood based on insufficient documentation it can cause unnecessary downtime.

Well, using such features usually causes unexpected and unnecessary downtime anyway. It is always a tradeoff and it should be used only in well-understood environments completely under the owner's control, and it that case it can probably be tested as well.
I remember in the early days of my DSL connection it was kind of unreliable and I made very complicated recovery scripts that had to analyze the situation to know what exactly was wrong.
This used similar mechanisms (pinging the modem, pinging the nexthop at the ISP, pinging something further down the internet) so cause actions like rebooting the modem, resetting the PPP
connection, or even switching to an ISDN dialup as a fallback.
Unfortunately it happened several times that this mechanism misfired due to silly things like the ISP changing the network layout so the nexthop changed, the routers at the ISP being reconfigured to no longer reply to ping, etc.
So you always need to have a watchful eye!

Furthermore, the pinging of a single IP is probably a bit too limited. You would want to ping at least in 2 directions (a system on the LAN and a system on the internet).
Netwatch is more versatile in that, because you can add more systems to watch, but unfortunately it does not have that "alert when 6 pings are missed" capability so when there is some ping loss it alerts all the time. It should be enhanced with a "number of pings to fail/succeed before state change is reported" field.
(which I consider more clear than the way it is implemented in watchdog)