New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 635901 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Aug 2016
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug



Sign in to add a comment

Exclude V8-specific values from System Health Perf Sheriffing alerts

Project Member Reported by petrcermak@chromium.org, Aug 9 2016

Issue description

Please exclude V8-specific values from the perf sheriffing shift set up in issue 627758.

In particular, this means changing the following monitoring path:

*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:*:effective_size_avg/*/*
                                                                          ^
                                                                          |

so that the fourth wildcard cannot contain colon. The following values should thus still be monitored:

memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg [wildcard matches 'v8']
memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg [wildcard matches 'malloc']
memory:chrome:all_processes:reported_by_chrome:partition_alloc:effective_size_avg [wildcard matches 'partition_alloc']
...

while the following values should NOT be monitored anymore:

memory:chrome:all_processes:reported_by_chrome:v8:heap:effective_size_avg [wildcard matches 'v8:heap']
memory:chrome:all_processes:reported_by_chrome:v8:heap:code_space:effective_size_avg [wildcard matches 'v8:heap:code_space']
...

Sample alert that should NOT be sent to the System Health perf sheriffing shift anymore: https://chromeperf.appspot.com/group_report?bug_id=635485.
 
Ulan: There are quite a few V8-specific alerts in https://chromeperf.appspot.com/alerts?sheriff=System%20Health%20Perf%20Sheriff. Do you want to triage them, or should I marked them as ignored?
Sorry I'm so slow on this! Our monitoring isn't fine-tuned enough to be a full regular expression. Would it be okay to type out all the values individually that currently would match the '*' not containing a colon? Or do we expect those to change over time?
Cc: primiano@chromium.org perezju@chromium.org
It's not ideal because it won't automatically start tracking newly added memory dump providers, but if necessary, we can do that:

memory:chrome:all_processes:reported_by_chrome:blink_gc:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:cc:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:discardable:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:dom_storage:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:font_caches:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:gpu:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:java_heap:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:leveldb:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:media:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:partition_alloc:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:sharedbitmap:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:skia:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:sqlite:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:tracing:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:web_cache:effective_size_avg
memory:chrome:all_processes:reported_by_chrome:winheap:effective_size_avg

The full paths are:

*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:blink_gc:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:cc:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:discardable:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:dom_storage:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:font_caches:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:gpu:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:gpumemorybuffer:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:java_heap:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:leveldb:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:media:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:partition_alloc:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:sharedbitmap:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:skia:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:sqlite:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:tracing:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:v8:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:web_cache:effective_size_avg/*/*
*/*/system_health.memory_*/memory:chrome:all_processes:reported_by_chrome:winheap:effective_size_avg/*/*

In the long term, would it be possible to add support for full regular expressions? I might be wrong, but I suppose it shouldn't be too much work.
Status: Fixed (was: Assigned)
I updated the paths.

Long-term, I think the whole system is pretty brittle: the benchmark author needs to ensure somebody set a pattern on the dashboard. What I'd really like to see is things like owners and monitoring set in the benchmark and then passed into the dashboard the same way we pass up units, descriptions, etc.
+1 to "owners and monitoring set in the benchmark", that would be incredibly useful.

Sign in to add a comment