Automatic Visibility/Observability

Hello all,

putting higher load on my test system, I occasionally observe dropping of span information. By that I mean, that for comparable spans sometimes the whole information "is there" according to my instrumentation and sometimes not. 

In this context I wanted to ask: If information is dropped, is it completely dropped? For example: If I want to make a statistic over all sql calls, calculating an average. And in one span less sql calls are registered, so a lower total number is recorded - can I trust the overall time given for the sql calls as being associated to the lower amount of calls? (I do that by filtering in one span for technology and choosing my db technology).

Thanks in advance,

Thomas

Hi Thomas,

In my experience, those symptoms are related to our self-hosted edition and may indicate an issue with performance in your disks. If you have setup the backend based on the requirements then you should be safe, if not, this is an indicator that you need to have the right resources. Remember, we don't do sampling and we capture all.

If everything is OK on your side in terms of resources, I suggest to contact our IBM Support team for further assistance.

Regards,

Hello all,

putting higher load on my test system, I occasionally observe dropping of span information. By that I mean, that for comparable spans sometimes the whole information "is there" according to my instrumentation and sometimes not. 

In this context I wanted to ask: If information is dropped, is it completely dropped? For example: If I want to make a statistic over all sql calls, calculating an average. And in one span less sql calls are registered, so a lower total number is recorded - can I trust the overall time given for the sql calls as being associated to the lower amount of calls? (I do that by filtering in one span for technology and choosing my db technology).

Thanks in advance,

Thomas

Hello Israel,

I am very positive, we have a SAAS endpoint and not a self-hosted edition. I just did not find the endpoint on the UI.

But that is really not my concern here: Just assume, that there is information dropping. Maybe just like in an underprovisioned self-hosted edition. Can I trust the "reduced" information. So, e.g., only half of the sql calls are recorded, that I expect from the code, is the aggregated latency of thes sql calls correct and not referring to actually more.

Thanks for your answer!

Thomas

Hi Thomas,

In my experience, those symptoms are related to our self-hosted edition and may indicate an issue with performance in your disks. If you have setup the backend based on the requirements then you should be safe, if not, this is an indicator that you need to have the right resources. Remember, we don't do sampling and we capture all.

If everything is OK on your side in terms of resources, I suggest to contact our IBM Support team for further assistance.

Regards,

Hello all,

putting higher load on my test system, I occasionally observe dropping of span information. By that I mean, that for comparable spans sometimes the whole information "is there" according to my instrumentation and sometimes not. 

In this context I wanted to ask: If information is dropped, is it completely dropped? For example: If I want to make a statistic over all sql calls, calculating an average. And in one span less sql calls are registered, so a lower total number is recorded - can I trust the overall time given for the sql calls as being associated to the lower amount of calls? (I do that by filtering in one span for technology and choosing my db technology).

Thanks in advance,

Thomas