This message was deleted.
# k3sa
adamant-kite-43734
09/19/2023, 9:18 AMThis message was deleted.
l
late-needle-80860
09/19/2023, 9:22 AMHi @stale-addition-90512 ,
You should provide some more details.
• k3s version
• Os and os version
• What database are you talking about? The db for etcd?
Spill the beans if you want effective help.
🙇♂️ 1
s
stale-addition-90512
09/19/2023, 9:23 AMthanks for advise!
l
late-needle-80860
09/19/2023, 9:23 AMOf course. No worries …. Happy to help
🙏 1
s
stale-addition-90512
09/19/2023, 9:24 AMk3s -v
k3s version v1.25.5+k3s2 (de654222)
go version go1.19.4
cat /etc/*release
NAME=“SLES”
VERSION=“15-SP3"
VERSION_ID=“15.3”
PRETTY_NAME=“SUSE Linux Enterprise Server 15 SP3"
ID=“sles”
ID_LIKE=“suse”
ANSI_COLOR=“0;32”
CPE_NAME=“cpe/osusesles15:sp3"
DOCUMENTATION_URL=“https://documentation.suse.com/”
and yes, db for etcd, i’m using postgresql
l
late-needle-80860
09/19/2023, 9:25 AMIt’s usually the disk iops at play in these scenarios. What’s your postgresql disk setup?
s
stale-addition-90512
09/19/2023, 9:26 AMit’s ssd
l
late-needle-80860
09/19/2023, 9:28 AM6GB NOT that much throughput … but could be enough depending on the pressure on the api.
RAID setup?
s
stale-addition-90512
09/19/2023, 9:28 AMk3s=#
select (blk_read_time+blk_write_time) AS iotime from pg_stat_statements order by (blk_read_time+blk_write_time) desc limit 5;
iotime
---------------------
203.53721000000058
15.213267
3.018391
0.21515399999999998
0.084551
(5 rows)
k3s=#
l
late-needle-80860
09/19/2023, 9:29 AMAlso backbone on the network side … what’s the setup?
s
stale-addition-90512
09/19/2023, 9:29 AMyes
stale-addition-90512
09/19/2023, 9:29 AMit’s localhost
l
late-needle-80860
09/19/2023, 9:29 AMRaid what?
s
stale-addition-90512
09/19/2023, 9:29 AMi think raid0
l
late-needle-80860
09/19/2023, 9:29 AMIsolation of disks for PostgreSQL and the rest of the kubernetes cluster?
s
stale-addition-90512
09/19/2023, 9:30 AMnope, but some ioweight setting
l
late-needle-80860
09/19/2023, 9:30 AMRegular workload on the same nodes as the api? And where the PostgreSQL is?
s
stale-addition-90512
09/19/2023, 9:30 AMyes, all workloads are one the some node, including DB
l
late-needle-80860
09/19/2023, 9:31 AMSo only one node? A one node cluster?
s
stale-addition-90512
09/19/2023, 9:31 AMyes, one node
l
late-needle-80860
09/19/2023, 9:31 AM🤠 🤠 🤠 🤠
late-needle-80860
09/19/2023, 9:31 AMAnd node type? Architecture and pc type
s
stale-addition-90512
09/19/2023, 9:32 AMit’s VEP-4600-V930
stale-addition-90512
09/19/2023, 9:33 AMmain question is where does this 629ms number come from?
l
late-needle-80860
09/19/2023, 9:34 AMSome VNF server … dedicated to network functions … no wonder this does not give the needed performance I would say
s
stale-addition-90512
09/19/2023, 9:35 AMwhat kind of performance we are talking about here? any specific requirement ?
l
late-needle-80860
09/19/2023, 9:35 AMI would think that the 629ms trace warning is from the api .. so that’s why you’re not seeing this in the dev select statement you did
late-needle-80860
09/19/2023, 9:36 AMIt’s the entire processing time spend from request to completion by whatever via the kubernetes api
s
stale-addition-90512
09/19/2023, 9:36 AMok, so + network delay
🎯 1
l
late-needle-80860
09/19/2023, 9:36 AMYou can read about the performance requirements on the etcd website in the docs
👍 1
late-needle-80860
09/19/2023, 9:37 AMEtcd is picky as it should be
s
stale-addition-90512
09/19/2023, 9:37 AMbut..i’m not really using etcd, i’m using postgres
l
late-needle-80860
09/19/2023, 9:37 AMOh yes sorry…
late-needle-80860
09/19/2023, 9:38 AMStill … the api is picky 🍕 🤡
late-needle-80860
09/19/2023, 9:39 AMI would say too much on one node .. it’s also possible to look requirements of the ….. api up on the kubernetes docs
🙌 1
2 Views