Description
Details
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | BUG REPORT | • nfraison | T325809 Presto is unstable with more than 5 worker nodes | ||
Resolved | • nfraison | T329525 Create Presto test clusters with 10 new nodes and try reproduce issue | |||
Resolved | MoritzMuehlenhoff | T329831 KDC performance tuning for TCP requests |
Event Timeline
Change 889971 had a related patch set uploaded (by Muehlenhoff; author: Muehlenhoff):
[operations/puppet@production] Tweak scalability of KDC requests
Change 889971 merged by Muehlenhoff:
[operations/puppet@production] Tweak scalability of KDC requests
Change 890388 had a related patch set uploaded (by Muehlenhoff; author: Muehlenhoff):
[operations/puppet@production] Revert "Tweak scalability of KDC requests"
Change 890388 merged by Muehlenhoff:
[operations/puppet@production] Revert "Tweak scalability of KDC requests"
Change 890389 had a related patch set uploaded (by Muehlenhoff; author: Muehlenhoff):
[operations/puppet@production] Tweak scalability of KDC requests (v2)
Change 890389 merged by Muehlenhoff:
[operations/puppet@production] Tweak scalability of KDC requests (v2)
Change 890393 had a related patch set uploaded (by Muehlenhoff; author: Muehlenhoff):
[operations/puppet@production] Adapt KDC monitoring to dynamic KDC worker count
Change 890393 merged by Muehlenhoff:
[operations/puppet@production] Adapt KDC monitoring to dynamic KDC worker count
The amount of KDC workers is now configurable via the new profile::kerberos::kdc::workers Hiera setting and has been raised to 8. In the addition net.core.somaxconn sysctl was bumped to 16k. This fixes the request burst issues observed in T329525
Change 891310 had a related patch set uploaded (by Muehlenhoff; author: Muehlenhoff):
[operations/puppet@production] Adjust monitoring for KDC processes if worker threads are in use
Change 891310 merged by Muehlenhoff:
[operations/puppet@production] Adjust monitoring for KDC processes if worker threads are in use