router: per-cluster histograms w/ timeout budget#9227
router: per-cluster histograms w/ timeout budget#9227mattklein123 merged 42 commits intoenvoyproxy:masterfrom
Conversation
This change adds a pair of histograms to track timeout budget usage: one for the per-request timeout and one for the global timeout. The histograms are scaled such that 100% of the timeout budget is a value of 10000 so that four digits of the percentage used can be recorded. Risk Level: Low (new stats only). Testing: Unit tests added/modified. Docs Changes: Added notes about historgrams. Release Notes: Added comment with link to docs. Fixes envoyproxy#6122 Signed-off-by: Matthew Gumport <[email protected]>
tonya11en
left a comment
There was a problem hiding this comment.
Thanks for getting the ball rolling on this! I mostly have a few questions/nits this pass.
docs/root/configuration/upstream/cluster_manager/cluster_stats.rst
Outdated
Show resolved
Hide resolved
docs/root/configuration/upstream/cluster_manager/cluster_stats.rst
Outdated
Show resolved
Hide resolved
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
gave up on that approach and should've gotten them before the review. Signed-off-by: Matthew Gumport <[email protected]>
personally hurt that this is part of ci Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
I'm going to agree with ci. Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
tonya11en
left a comment
There was a problem hiding this comment.
LG in general. Just one final set of comments and then we can rope in the maintainers.
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
Signed-off-by: Matthew Gumport <[email protected]>
tonya11en
left a comment
There was a problem hiding this comment.
LGTM modulo the nit, but that's your call. I think a maintainer should take a look now.
|
Also, can you write a bit about the flag you added that toggles the histograms in the description. |
mattklein123
left a comment
There was a problem hiding this comment.
Nice! LGTM with some small comments. Thank you.
/wait
docs/root/configuration/upstream/cluster_manager/cluster_stats.rst
Outdated
Show resolved
Hide resolved
Signed-off-by: Matthew Gumport <[email protected]>
|
Going to re-merge master for the stats tests. /wait |
…-stat Signed-off-by: Matthew Gumport <[email protected]>
|
Still waiting for new values. /wait |
Signed-off-by: Matthew Gumport <[email protected]>
|
Going to merge master again. |
…-stat Signed-off-by: Matthew Gumport <[email protected]>
|
/wait |
Signed-off-by: Matthew Gumport <[email protected]>
|
Test integration foobar /wait |
Signed-off-by: Matthew Gumport <[email protected]>
mattklein123
left a comment
There was a problem hiding this comment.
LGTM modulo merge issues. Nice work!
/wait
Signed-off-by: Matthew Gumport <[email protected]>
This change adds a pair of histograms to track timeout budget usage: one for the
per-request timeout and one for the global timeout. The histograms are scaled
such that 100% of the timeout budget used (a timeout) is a value of 100.
The histograms by default are off. They can be turned on by setting
track_timeout_budgetson a cluster to true.Risk Level: Low (new stats only).
Testing: Unit tests added/modified.
Docs Changes: Added notes about histograms.
Release Notes: Added comment with link to docs.
Fixes #6122
Signed-off-by: Matthew Gumport [email protected]