xotcl-request-monitor

  • last updated 17 hours ago
Constraints
Constraints: committers
 
Constraints: files
Constraints: dates
drop FK on community_id for cases, where communities are frequently dropped.

we do not want to loose statistics in these cases

file upgrade-0.60-0.61.sql was initially added on branch oacs-5-10.

    • -0
    • +0
    ./sql/postgresql/upgrade/upgrade-0.60-0.61.sql
reduce verbosity (set verbose_blocking per default to 0), improve speed of dump creation

New method to cleanup all data associated with a community in the throttle object:

if a community is deleted when a user is accessing it, some leftover information will be kept in the object. At the next request this will be interpreted as the community having been left and the system will try to track this event into request_monitor_community_activities. Depending on who comes first, we might try to insert a tuple in the table for an object that does not exist anymore.

Hence, we use this method to cleanup all community info before the actual deletion in dotlrn (happening only downstream).

request monitor overhaul:

- new parameter "do_double_click_prevention" (default: on)

When turned off, double click prevention can be selectively

turned off.

- new parameter "monitor_urls" (default: / /register/ /dotlrn/)

track for these URLs performance via minin responsetime plugin

- added upgrade script for removing old parameters

- bump version to 0.60

file callback-proc.tcl was initially added on branch oacs-5-10.

    • -0
    • +0
    ./tcl/callback-proc.tcl
let watchdog send mail to webmaster with data of active requests

pool_remap_watchdog: new watchdog for potentially stuck processes

This watchdog checks every minute the running jobs and maps very slow

requests to the slow pool (if configured) to avoid that the

default pool is getting filled up with more stuck requests.

This ensures the livelyness of a server.

reduce verbosity

Fix typo

Update proctoring upload URL to the new one

send also HEAD requests to the request monitor

fix filtering logic

prefer "ns_conn isconnected" over "ad_conn isconnected", since the latter might not notice changes in the connect state

  1. … 18 more files in changeset.
avoid double loading of bootstrap

added pool to long-calls statistics (requires at least NaviServer 4.99.5, May 2013)

use "ad_file" as a means to avoid unexpected tilde substitution in file names

improve spelling: move closer to the linux documentation recommendations

  1. … 34 more files in changeset.
fix typos

add more query details to logged query

add entries to longcalls as well when filtertime > 1s, or queuetime > 0.5s

make on/off logic less confusing

make monitor URLs easier configurable

don't use request-blocking on embedded items

improve performance of perDayCleanup significantly by not persisting Users.hits

remove all of /shared from request blocking

make blocked request easily detectable in monitoring tools like elasticsearch by using 429 as HTTP result code

added connection pool management

TODO: improved parmeterization

fix variable name (many thanks to Franz Penz)

whitespace cleanup

add support for deleted users, handle this case centrally