Incorrect bandwidth calculations from dual homed servers #5695
Assignees
Labels
Milestone
Comments
|
I'll take this one on! |
tcfdev
pushed a commit
to tcfdev/trafficcontrol
that referenced
this issue
Apr 1, 2021
Per bug report apache#5695, any and all interfaces returned in polling were used to calculate vitals, such as bytes in or out, bandwidth and more. However, only designated interfaces should be used for calculating such metrics and vitals. As such, this commit will check for monitored interfaces and only use those for the calculations.
6 tasks
ocket8888
pushed a commit
that referenced
this issue
Apr 15, 2021
…uded in the vitals (bandwidth) (#5730) * Normalizing README.md for MacOS case insensitive discrepancies * Removed whitespace at end of line * Only use monitored interfaces for vitals Per bug report #5695, any and all interfaces returned in polling were used to calculate vitals, such as bytes in or out, bandwidth and more. However, only designated interfaces should be used for calculating such metrics and vitals. As such, this commit will check for monitored interfaces and only use those for the calculations. * Normalized README.md file * Normalized README file * Add bugfix info to CHANGELOG * Check monitor config for nil Ensure callers for GetVitals must include a valid config and the config should contain interfaces * Fix merge issue with CHANGELOG * Address PR feedback. Remove extra blank line from CHANGELOG Condense test structs for sake of clarity Change log levels to reflect actual logging issues Add logging clarity to missing or empty caches and interfaces Co-authored-by: Taylor Frey <[email protected]>
ocket8888
pushed a commit
that referenced
this issue
Apr 26, 2021
…uded in the vitals (bandwidth) (#5730) * Normalizing README.md for MacOS case insensitive discrepancies * Removed whitespace at end of line * Only use monitored interfaces for vitals Per bug report #5695, any and all interfaces returned in polling were used to calculate vitals, such as bytes in or out, bandwidth and more. However, only designated interfaces should be used for calculating such metrics and vitals. As such, this commit will check for monitored interfaces and only use those for the calculations. * Normalized README.md file * Normalized README file * Add bugfix info to CHANGELOG * Check monitor config for nil Ensure callers for GetVitals must include a valid config and the config should contain interfaces * Fix merge issue with CHANGELOG * Address PR feedback. Remove extra blank line from CHANGELOG Condense test structs for sake of clarity Change log levels to reflect actual logging issues Add logging clarity to missing or empty caches and interfaces Co-authored-by: Taylor Frey <[email protected]> (cherry picked from commit 89669fd)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
I'm submitting a ...
Traffic Control components affected ...
Current behavior:
Traffic Monitor indiscrimantly sums up bandwidth calculations from all reported interfaces, whether the interface is marked as "Monitor this interface" or not. When more than 1 interface is returned for a cache server it is assumed to be part of the calculations. This results in inaccurate numbers as some interfaces are "nested" / children interfaces or loopback which result in 2x calculations at a minimum.
Expected behavior:
Only the interfaces marked with "Monitor this interface" should be included in the calculations.
Minimal reproduction of the problem with instructions:
Query a Traffic Server cache where the
/_statsendpoint returns multiple interfaces, including those that may not necessarily be noted in the TP/TO. These additional interfaces will be included in the bandwidth calculation that can then be seen in the Traffic Monitor endpoint/CacheStatsor/CacheStatsNew. The bandwidth should be based only on the interfaces marked as "Monitor this interface" as can be seen in the Traffic Portal web application.The text was updated successfully, but these errors were encountered: