Re: [atlas] Email or SMS alert when probe goes offline/online
On Wed, 7 Sep 2011 12:35:42 +1000 "Robert" <robert@robert.id.au> wrote:
What do other participants think about having email alerts when a probe goes offline or online?
Wasn't something like this implemented just recently? --------------------------------------------------------------------------------------------------------- From: Robert Kisteleki <robert@ripe.net> To: ripe-atlas@ripe.net Subject: [atlas]Atlas current news Date: Thu, 14 Jul 2011 14:46:26 +0200 Sender: ripe-atlas-admin@ripe.net User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:5.0) Gecko/20110624 Thunderbird/5.0 Organization: RIPE NCC Dear Atlas users, We've added two new features to RIPE Atlas recently: 1. Automatic notification about down probes: if your probe has been down for some time, you'll get a one-time automatic notification. This happens after five consecutive days of downtime. 2. Static network configuration: after the testing round with volunteers, we decided to make this feature publicly available, thus released the 4.030 firmware version. You can read more about this feature at https://atlas.ripe.net/doc/static-config Regards, Robert Kisteleki, for the RIPE Atlas Team -- With respect, Roman
On 9/7/11 9:43 AM, Roman Mamedov wrote:
We've added two new features to RIPE Atlas recently:
1. Automatic notification about down probes: if your probe has been down for some time, you'll get a one-time automatic notification. This happens after five consecutive days of downtime.
Right, and I personally would prefer that to be much sooner. Say after a couple of hours or so. Maybe it would be nice to make it configurable by the host. Bert
On 07.09 09:57, Bert Wijnen wrote:
On 9/7/11 9:43 AM, Roman Mamedov wrote:
We've added two new features to RIPE Atlas recently:
1. Automatic notification about down probes: if your probe has been down for some time, you'll get a one-time automatic notification. This happens after five consecutive days of downtime.
Right, and I personally would prefer that to be much sooner. Say after a couple of hours or so. Maybe it would be nice to make it configurable by the host.
Indeed. There are two things here: 1) We want to re-activate probes that are down due to local problems, e.g. problems that the host can do something about. This is in the interest of the whole community because it helps to keep a large measurement network up and operating. For this purose we are already sending out e-mails after a couple of days of downtime. This is not going to change and we are considering more 'agressive' poking if the e-mails do not bring the probe back up. 2) Providing a service to the hosts that warns them of downtime etc. This should be configurable of course in terms of who gets alerted and how sensitive the trigger for an alert is; the default shold be "off". I have the feeling that this is also someting we may want to address as part of a more general alerting mechanism. As a host I might be interested in monitoring a totally different target, my family server located at a hoster, and do it not only from my own probe. That is called user defined measurements and an alerting system based on that will cover the case we are discussing now, because one will be able to monitor the first upstream from one's own probe. Right? Daniel
Hi, personally i would really like such a feature. And from the responses so far i think the mechanism should account for the different mindset of people hosting a probe. I agree with Daniel that the default should be off and the sensitivity should be configurable (e.g. alert after X time offline):
2) Providing a service to the hosts that warns them of downtime etc. This should be configurable of course in terms of who gets alerted and how sensitive the trigger for an alert is; the default shold be "off".
Maybe i can extend a bit on this proposal. E.g. with the possibility to specify an alternate email address (other than for the registration) these alert mails could automatically be filtered. Personally i would prefer a mechanism that alerts me as soon as connectivity is lost for more than X minutes, with 1 < X < 10. In case i would get to many of these notifications, i would just go to the webpage and turnoff the notifications. Another option that just came to my mind is to send an activity report every month or so. This could then include a fraction of time the probe was up and/or connectivity was present. Maybe with a link to the status webpage. So that people are reminded of the existence of the probe once in a while. (Of course this should again be an opt-in mechanism. I would certainly subscribe.) best Fabian
Hello,
Maybe i can extend a bit on this proposal. E.g. with the possibility to specify an alternate email address (other than for the registration) these alert mails could automatically be filtered.
Personally i would prefer a mechanism that alerts me as soon as connectivity is lost for more than X minutes, with 1 < X < 10. In case i would get to many of these notifications, i would just go to the webpage and turnoff the notifications.
This is doable and there seems to be demand for it, so we'll make it happen. Currently we're working on something bigger (user defined measurements), but we'll remember your words :-)
Another option that just came to my mind is to send an activity report every month or so. This could then include a fraction of time the probe was up and/or connectivity was present. Maybe with a link to the status webpage. So that people are reminded of the existence of the probe once in a while. (Of course this should again be an opt-in mechanism. I would certainly subscribe.)
Virtually all the information you mention is available on the probe status page already, but indeed we could introduce a mechanism to send reports to the host. Let us know if you'd like such a feature. Regards, Robert
best Fabian
Robert Kisteleki wrote:
Hello, [...]
Another option that just came to my mind is to send an activity report every month or so. This could then include a fraction of time the probe was up and/or connectivity was present. Maybe with a link to the status webpage. So that people are reminded of the existence of the probe once in a while. (Of course this should again be an opt-in mechanism. I would certainly subscribe.)
Virtually all the information you mention is available on the probe status page already,
Correct. Also the possibility to download the "raw data" is great, many thanks for that!
but indeed we could introduce a mechanism to send reports to the host. Let us know if you'd like such a feature.
This actually led me to changing the subject line. Right now consumption of the data and results on the probe owners' ends is based on "pull". Which is a Good Thing [TM] in itself. However, without doing extensive and error-prone scripting, it is difficult to automate and to feed into exisiting local machinery. So - here is a plee for coming up with some framework that allows me to have (selected .or. all) data *pushed* to a target on my end. This could be as simple as sending that stuff to a mailbox I vreate for that purpose and which I can register with the data repoitory hub. Of course, some safeguards apply to prevent from people "data-bombing" innocent recipients :-) This would also support sharing the data for my probe(s) with other parties for comparison - at least as long as there isn't a more fine-grained access permission system in place at the hub than [private .xor. fully-public]. Any other distribution technology that may be appropriate is fine, of course. I would immediately use that service to collect and archive "my data" on a regular basis!
Regards, Robert
best Fabian
Thanks, Wilfried.
participants (6)
-
Bert Wijnen -
Daniel Karrenberg -
Fabian Schneider -
Robert Kisteleki -
Roman Mamedov -
Wilfried Woeber, UniVie/ACOnet