Sentry User Feedback

[pmcross]

When I said that false alerts made it through, I meant that sms messages were sent for every single alert clip, even without people in it and clips with people had no logo except for the few occupied marks on about 1 out of 10 clips. The sentry had been working correctly i days prior, and nothing was changed on my end however the sentry log entries stopped appearing when all alert clips were sending sms messages. The pass all clips problem however seems to have resolved between 10:22 and 10:50pm CST last night as I received no SMS overnight, and the motion entries are followed by the sentry stat counter messages in the logs. Also the sentry log entries follow the motion entries by 3-5 seconds where yesterday i saw that delay grow to over 1.5 hours at one point. I will update as soon as I verify person detection and logos are working again.

This is what I meant by am saying false positives made it through and weren’t suppressed. I apologize for the confusion. Also, I will contact support regarding verifying the trial on the rest of my cameras. I will verify that the problem is resolved as well, but I believe that it is as I stopped receiving alerts last night that didn’t contain a person. Thanks again!


Sent from my iPhone using Tapatalk

 

[varghesesa]

Great feedback. Thank you! Thank you! We will continue to investigate internally. I think you found an issue.

Original delay of 3-5 sec before clips get marked sounds right. That is similar to the behavior on my installation. Something was not right.

Sam from Sentry

Engineering reviewed the traffic on 2019-Mar-30. We did hit some capacity issues from 6:05 am PDT - 1:05 pm PDT. BI would timeout during this time and send the alert to error on the safe side.

This is not to say the system was down but requests on occasion would be dropped or timeout and BI would send the alert which is consistent to the experience described by many where all alerts were being sent.

Sorry for inconvenience.

-- Sam from Sentry

 

[pmcross]

Engineering reviewed the traffic on 2019-Mar-30. We did hit some capacity issues from 6:05 am PDT - 1:05 pm PDT. BI would timeout during this time and send the alert to error on the safe side.

This is not to say the system was down but requests on occasion would be dropped or timeout and BI would send the alert which is consistent to the experience described by many where all alerts were being sent.

Sorry for inconvenience.

-- Sam from Sentry

Thanks for the update, Sam. Greatly appreciated!


Sent from my iPhone using Tapatalk

 

[integlikewhoa]

Engineering reviewed the traffic on 2019-Mar-30. We did hit some capacity issues from 6:05 am PDT - 1:05 pm PDT. BI would timeout during this time and send the alert to error on the safe side.

This is not to say the system was down but requests on occasion would be dropped or timeout and BI would send the alert which is consistent to the experience described by many where all alerts were being sent.

Sorry for inconvenience.

-- Sam from Sentry

Sam,

We appear to be having this same problem again. BI is falsely triggering and the BI logs are showing "Sentry: 15s timeout"

Thanks,
JIM

 

[Matias G Labarrere]

Sam,

We appear to be having this same problem again. BI is falsely triggering and the BI logs are showing "Sentry: 15s timeout"

Thanks,
JIM

I’ve gotten several this am - same result


Sent from my iPhone using Tapatalk

 

[pmcross]

I have several of these in my BI logs today too.

 

[varghesesa]

Sam,

We appear to be having this same problem again. BI is falsely triggering and the BI logs are showing "Sentry: 15s timeout"

Thanks,
JIM

You guys are one step ahead of me. Not home right now to test today's update. Spoke with BI this morning and we discussed ways to improve the logs to provide more meaningful messages to users as to what is going on with the integration.

We had suspicions that BI is timing out before Sentry is able to respond, thus the new log messages to get confirmation.

Scaling is still a problem obviously. Letting engineering know.

-- Sam from Sentry.

 

[varghesesa]

FYI, to be conservative, BI sends an alert when it times out so no alerts are missed. Better safe than sorry.

-- Sam from Sentry

 

[integlikewhoa]

FYI, to be conservative, BI sends an alert when it times out so no alerts are missed. Better safe than sorry.

-- Sam from Sentry

I'm just curious how big of a problem this is and if you will be able to scale enough and be affordable. Right now you don't have many customers and we are still running into server overload issues. Failsafe of the 15sec time out is good but even if the servers respond in 10sec that is still slow for me. I have automation based on motion detection. Walking up to a front door at night turns on a light and other things. What used to be local detection and fast is now turning into a 10-15sec delay. And when you standing on a porch 15sec seems like 5 min. I also have other things going on that take snaps shots when motion is detected and now 10sec means the person as left the fame. So all this delay is killing me where its not worth the hassle. Detection should be better but I still get false negatives and positives and I have no control over them any more. I'm hoping this not only stops timing out but rather gets faster like under 5 sec tops and i would prefer faster but I know i can't ask for to much.

Also we need multiple frame upload to help with false reports but if a single frame is causing this overload already I can only imagine why were are not running 3-4 frames.

So I ask how big of a problem is this behind the scenes and how realistic is it $$$ to fix.

 

[fenderman]

I'm just curious how big of a problem this is and if you will be able to scale enough and be affordable. Right now you don't have many customers and we are still running into server overload issues. Failsafe of the 15sec time out is good but even if the servers respond in 10sec that is still slow for me. I have automation based on motion detection. Walking up to a front door at night turns on a light and other things. What used to be local detection and fast is now turning into a 10-15sec delay. And when you standing on a porch 15sec seems like 5 min. I also have other things going on that take snaps shots when motion is detected and now 10sec means the person as left the fame. So all this delay is killing me where its not worth the hassle. Detection should be better but I still get false negatives and positives and I have no control over them any more. I'm hoping this not only stops timing out but rather gets faster like under 5 sec tops and i would prefer faster but I know i can't ask for to much.

Also we need multiple frame upload to help with false reports but if a single frame is causing this overload already I can only imagine why were are not running 3-4 frames.

So I ask how big of a problem is this behind the scenes and how realistic is it $$$ to fix.

I always think its funny when users try to question business models and make assumptions to come to silly conclusions. You have no idea how many customers they have as there is a free trial. You have no idea how many blue iris users signed up for the free trial. Do you think they simply forgot about the bandwidth and processing needs to scale? Its 2019, you can easily scale the processing and bandwidth needs. This is a NEW integration it takes time to work out, that is why there is a free trial. If its not worth the hassle for you then move on. 15 seconds is perfectly acceptable for an alert time.

 

[pmcross]

You guys are one step ahead of me. Not home right now to test today's update. Spoke with BI this morning and we discussed ways to improve the logs to provide more meaningful messages to users as to what is going on with the integration.

We had suspicions that BI is timing out before Sentry is able to respond, thus the new log messages to get confirmation.

Scaling is still a problem obviously. Letting engineering know.

-- Sam from Sentry.

Thanks for the update, Sam. Are you referring to the BI update that was released today or a Sentry update? Definitely understand about scaling. They are just growing pains that will work their selves out

 

[varghesesa]

I'm just curious how big of a problem this is and if you will be able to scale enough and be affordable. Right now you don't have many customers and we are still running into server overload issues. Failsafe of the 15sec time out is good but even if the servers respond in 10sec that is still slow for me. I have automation based on motion detection. Walking up to a front door at night turns on a light and other things. What used to be local detection and fast is now turning into a 10-15sec delay. And when you standing on a porch 15sec seems like 5 min. I also have other things going on that take snaps shots when motion is detected and now 10sec means the person as left the fame. So all this delay is killing me where its not worth the hassle. Detection should be better but I still get false negatives and positives and I have no control over them any more. I'm hoping this not only stops timing out but rather gets faster like under 5 sec tops and i would prefer faster but I know i can't ask for to much.

Also we need multiple frame upload to help with false reports but if a single frame is causing this overload already I can only imagine why were are not running 3-4 frames.

So I ask how big of a problem is this behind the scenes and how realistic is it $$$ to fix.

We are on AWS so scaling is normally trivial, but for some reason our architecture is capping incoming requests. Engineering is getting closer to figuring out this issue. Sorry.

Agreed, 10-15 sec is slow. Normal behavior is 3-5 sec which we also want to get down to be able to address your chained events.

We are very focused on false negatives. Your suggestion of taking more frames as turnaround time to BI improves is a simple fix but we have other ideas as well. Improvements to come.

Sorry for the current frustration. Better days ahead.

-- Sam from Sentry

 

[varghesesa]

Thanks for the update, Sam. Are you referring to the BI update that was released today or a Sentry update? Definitely understand about scaling. They are just growing pains that will work their selves out

Correct, BI had an update yesterday to provide better info in the logs.

Thanks Phil for the support!

-- Sam from Sentry

 

[Matias G Labarrere]

Correct, BI had an update yesterday to provide better info in the logs.

Thanks Phil for the support!

-- Sam from Sentry

The BI update allows me to verify consistent timeouts throughout the day in my case, which is certainly helpful. I appreciate this update as it provides visibility as to which alerts are being pushed through vs properly scanned via sentry. In my instance, I’d say half of them are timing out. No need to respond just wanted to provide feedback for informational purposes.


Sent from my iPhone using Tapatalk

 

[integlikewhoa]

Sam,

We appear to be having this same problem again. BI is falsely triggering and the BI logs are showing "Sentry: 15s timeout"

Thanks,
JIM

We are on AWS so scaling is normally trivial, but for some reason our architecture is capping incoming requests. Engineering is getting closer to figuring out this issue. Sorry.
-- Sam from Sentry

I'm getting the same timeout messages again today.

 

[varghesesa]

Hi All,

According to engineering, we should be past the timeouts. If you are still receiving many Sentry timeoouts in the logs, please let us know. support@smarthomesentry.com

-- Sam from Sentry

 

[lifeatredline]

Hi All,

According to engineering, we should be past the timeouts. If you are still receiving many Sentry timeoouts in the logs, please let us know. support@smarthomesentry.com

-- Sam from Sentry

Today is the first day in awhile that Sentry has worked all day. It has been working fine at night, but no response or timeout in daytime. Looks like engineering is right so far. Very pleased with how things are running. Thanks Sam & team for the hard work and communications. Here's to continued success and smooth running.

 

[sorka]

FYI, to be conservative, BI sends an alert when it times out so no alerts are missed. Better safe than sorry.

-- Sam from Sentry

Does this mean you'd get a false positive if you're only expecting an SMS when a person is detected? Maybe this explains all of the false positives I started getting in the middle of the night. I had to turn it off for now until I can confirm that issue is fixed.

 

[integlikewhoa]

Does this mean you'd get a false positive if you're only expecting an SMS when a person is detected? Maybe this explains all of the false positives I started getting in the middle of the night. I had to turn it off for now until I can confirm that issue is fixed.

Yes, that can and was happening up until last Friday (4-5-19). You can tell if this is happening by looking at your logs in BI. Also look for the motion logo in BI alerts. If the logo is there it thinks it saw a person. If the logs say something about 15sec time out then it means that sentry never responded to BI and BI sent an alert anyways to be safe.

 

[sorka]

We really need better notification customization for different sources. Just like you can setup notifications for failed cameras, we need notifications for other failed source types so that we an either send those notifications or filter them where we receive them.

The current alerts don't have any meta data on what type of alert it is. Motion? Human? <other thing>? ...so I have not way of knowing from a notification received what it was.