Hi,

Alost every time after server restart, one of my specific ping sensor goes mad. Before restart it shows acceptable values like 0..10ms but sometimes, after I restarted the server hosting PRTG, this ping sensor shows values like 60...160ms.. I can quarantee that this is not true - I have alternative monitoring solution also set up, wich pings the same device and shows no anomalies. I'm suspicious, that after restart this specific ping sensor sends out queries at the same time as some other sensors are collecting data. I'm using some historical server to run PRTG.. Dell Poweredge 1850 with two Intel Xeon 3.2GHz and 8GB of RAM running on Windows 2008R2. How can I point out, that the problematic side is our server side not the device itself? System, Core and Probe healt are fine.. Only way to get rid of this problem is to play with specific sensor scanning interval. Any helpful ideas?

Article Comments

Hello,

thank you very much for your KB-Post. May I ask, can you please upload a few screenshots showing the issue, and the "Settings"-Tab of this very Ping-Sensor?

best regards.


Jan, 2014 - Permalink

Hi, Below You'll find 3 screenshots: 1. http://emu.standard.ee:81/PRTG/1.png - showing anomalies about 02.01 and 25.01 on this year. On 02.01 it just goes mad and at the moment I don't remember what I did exactly to fix this. Last time (25.01) I get this problem fixed by setting scanning interval to 5 minutes and after some time back. Also I restarted local probe. 2. http://emu.standard.ee:81/PRTG/2.png - image of settings tab. 3. http://emu.standard.ee:81/PRTG/3.png - image of sensor probe health

When I ping this device from command line on the same time on this server I see those anomalies too. When I ping this device from command line on the server next to it, I don't see those anomalies. This leads me to think that there is some congestion on this server at this very moment when it's pinging.


Jan, 2014 - Permalink

Are there other sensors on the same device? If yes, this could indeed be sign of load to a certain extent, when PRTG starts, because it tries to get all sensors started as quickly as possible.
You could maybe try consider deploying a Remote Probe on this target, which would then continue monitoring if the PRTG Server is restarted
Does this also happen if you only restart the PRTG Services (and do not reboot the entire machine)?


Jan, 2014 - Permalink

Hi,

Yes. There are other sensors on the same device. Four SNMP traffic sensors (60 sec. interval) and one Sensor Factory Sensor (60 sec.interval) and one SNMP uptime sensor (1 h interval). Remote probe for just this device is not reasonable right now... I must check - restarting only services. But if I remember correctly, this also happends once if I upgraded PRTG components without restarting server itself..


Jan, 2014 - Permalink

Maybe it changes something if you set the Ping sensor to a smaller interval (30 seconds), and/or setting the Ping Sensor to send multiple ping packets per scan.


Jan, 2014 - Permalink

Must try. I hav not changed it to a smaller.. Basically, if I have restarted the server and when I'm going to see anomalies, I'm going to play with those settings already to get things fixed. May I ask, how are sensors workload distributed? is there any logic behind that to avoid scanning constantly on the same time..? To avoid peaks like that I'm suspecting right now? Any way to change sensor startup time on the same device like T+ 38 sec. os smthng like that?


Jan, 2014 - Permalink

After a restart of PRTG, all sensors of one device scan in quick succession to get all sensors running quickly. Once they show results, the scans are evenly distributed according to the scanning interval(s) of the sensors, so that possibly no peaks appear.


Jan, 2014 - Permalink

Does this distribution works also between devices? I'm mean, what is the chance of overlapping, if I use like 10 device with ping sensor and they all are configured using 60 sec. inteval. How big is the possibility that let's say 4 devices will be pinged at the same time(frame) ?


Jan, 2014 - Permalink

The distribution is mostly done on a per-device basis. Distributing all sensors of all devices evenly is of course also attempted, but much harder due to the different scanning intervals. It's not possible to judge the chance of x devices being pinged at the same time, there is a huge random factor involved, as this will change with each run / start of the PRTG services.


Jan, 2014 - Permalink