在大规模场景下，降低agent对apiserver的访问压力 #253
Conversation
|
|
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #253 +/- ##
==========================================
- Coverage 32.35% 31.94% -0.42%
==========================================
Files 41 41
Lines 6426 6443 +17
==========================================
- Hits 2079 2058 -21
- Misses 4058 4096 +38
Partials 289 289
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
pkg/csi/nodeserver.go
Outdated
|
|
||
| time.Sleep(time.Millisecond * 100) | ||
| time.Sleep(time.Second * 60) |
There was a problem hiding this comment.
如果为60s，增加一块磁盘是不是也要1分钟才能生效？
如果是的话，改成1s也影响不大，毕竟都有差异化diff了
There was a problem hiding this comment.
如果为60s，增加一块磁盘是不是也要1分钟才能生效？ 如果是的话，改成1s也影响不大，毕竟都有差异化diff了
对大规模集群来说，1s感觉还是有点频繁，而且对于没有使用到spdk的场景来说就是无意义的检测。延长到3-5秒或改成时间间隔可配置，是否可以？
There was a problem hiding this comment.
如果为60s，增加一块磁盘是不是也要1分钟才能生效？ 如果是的话，改成1s也影响不大，毕竟都有差异化diff了
在1000节点规模的集群里，我们实测安装agent比不安装要多消耗10个核的cpu，改成1秒理论上也有1核cpu的消耗
There was a problem hiding this comment.
可以的改成3-5ok的
已加单元测试，帮再review下，谢谢
The checkSPDKSupport frequency is too high, which puts a lot of pressure on apiserver in large cluster
ypnuaa037
force-pushed
the
reduce-agent-access-to-apiserver
branch
from
9275828
to
6bc3046
Compare
【背景问题】
在大规模集群场景下，发现agent对apiserver的压力很大。通过分析发现，主要源于两处：
其中，问题1是主要压力来源
【解决思路】