Red Hat Bugzilla – Bug 1513139
Client side rate limiting makes migrations excessively long for large clusters
Last modified: 2018-04-05 05:32:46 EDT
Description of problem: `oadm migration authorization` is going to take over 5 hours for a single large cluster (~22000 projects). The majority of this time will be spent in rate limiting logic in the client (200ms for rate limit vs 10ms for the request itself). Version-Release number of selected component (if applicable): v3.6 How reproducible: 100% Expected results: Rate limiting should be a user choice and likely not the default for operations required for upgrades. A 5 hour pre-check is a sizable impediment to a predictable upgrade. Can we turn this off for migrate or make a general purpose --no-rate-limit CLI flag? If an authenticated user wants to intentionally DOS attack the server, they can do so without using the CLI. Only a well intentioned user would bother using the CLI and, if they know what they are doing and opt in with the flag, we should let them migrate quickly.
Tested the following clients against a 3.6.173.0.86 server: 3.6.173.0.101 3.7.27 3.8.32 3.9.0-0.36.0 For 1000 projects oc adm migrate authorization time was reduced from 14 minutes (pre-fix client - 3.6.173.0.86) to 6-7 seconds. The 14 minutes is inline with the estimate above of 5 hours for 20K projects For 20000 projects, migration time with the fix was ~ 2m 20s for all versions of the fixed client. This can be marked VERIFIED once a functional team regression of the command is performed.
Verified,no issue found.
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2018:0636