Datasource Statistics may obtain invalid data for some rebooted devices

[jdcoats]

I have been finding that from time to time dsstats will get invalid data and have to be purged. I have finally found a way to reproduce this issue on demand. Its not always this device or data type but here is an example of what happens when I reboot this Linux server.

2021/12/08 10:25:14 - CMDPHP intropage-hr Backtrace: (/plugins/intropage/intropage.php[39]:process_page_request_variables(), /plugins/intropage/include/functions.php[87]:intropage_detail_panel(), /plugins/intropage/include/functions.php[594]:busiest_traffic_detail(), /plugins/intropage/panellib/busiest.php[1143]:human_readable(), /plugins/intropage/include/functions.php[1579]:cacti_debug_backtrace())
--
2021/12/08 10:25:14 - CMDPHP INTROPAGE WARNING: Bytes = [4.919131751469E+18], Factor = [1024], i = [6] d = [1152921504606846976]

[TheWitness]

Just finishing this up today on my day off.

[TheWitness]

@jdcoats, you need to update lib/dsstats.php, poller_dsstats.php and include/global_settings.php. After that see if you can reproduce again. Make sure the device has been up for a while and has a lot of traffic.

This change will implement RRDtool like clipping for COUNTERS and DERIVE type objects. If the resulting value is above the rrd_maximum value, it'll return NULL for the delta. A small gap will appear in the data, but that's fine. Maybe I can improve that. Not today.

It also addresses #4500, which will allow DSStats to finish sooner. Warning, the PHP processes take a lot of CPU.

[jdcoats]

Thanks! No longer able to reproduce on demand.

[TheWitness]

Cool. Ignore my other comment. The new parallel setting should speed things up too.

[TheWitness]

Here is without boot.

Here is with boost.

[jdcoats]

[TheWitness]

Cool. You should try like 4 processes and see if it's better.