Search results

Write back cache, reduce network latency. And read the testing and optimization guide,tests, that's been done by others and published in the forum

not a proxmox issue,just a general dhcp issue. Running a windows DC with dhcp, experiencing dhcp issues, clients not getting Ips taking long time etc. ran perfDHCP against the server and this is the result Running: perfdhcp -x i -t 2 -r 10 -R 100 10.a.b.c Scenario: basic. Multi-thread mode...

I really think proxmox would gain alot by having a GUI Wizard for importing OVA files. I know it's easy in CLI, but all the new customers are comming from VMware and but not adding such a small and simple feature lots of new customers are turning away. Considering the current state or vmware...

That is slow seq read/write. But CEPH is really slow, you need to adhere to the standards, ie five nodes with multiple OSDs per node. Im new to CEPH myself, but my limited testings yields similar results, but not that bad. I have only used very olh HGST 200GB sas ssd, 6xOSD per node in a 3 node...

I agree! some basic functions need to be in place for operations

Thank you for the information. You are correct # ceph osd pool autoscale-status POOL SIZE TARGET SIZE RATE RAW CAPACITY RATIO TARGET RATIO EFFECTIVE RATIO BIAS PG_NUM NEW PG_NUM AUTOSCALE BULK .mgr 1985k 3.0 3353G 0.0000...

Recovery/ Rebalance What could cause this behaviour ? all disks are fine, no osd failing. One empty vm are using the ceph as storage so bascially no usage. I have verified smartctl on all ssds. why is the cluster rebalancing ? now and then ? is that expected behaviour?

lxc: Cbs: 16 (f=16): [m(16)][31.8%][r=2032KiB/s,w=368KiB/s][r=127,w=23 IOPS][eta 13m:51s] fio: terminating on signal 2 Jobs: 16 (f=16): [m(16)][31.8%][eta 13m:54s] randrw: (groupid=0, jobs=16): err= 0: pid=913: Wed Jan 10 22:15:24 2024 read: IOPS=3474...

Hm, what happened to my cpeh cluster ? ~# ceph -w cluster: id: 47a8ff1a-0599-4215-b268-b4c06ef9274e health: HEALTH_OK services: mon: 3 daemons, quorum pm3,pm1,pm2 (age 10h) mgr: pm3(active, since 10h), standbys: pm1, pm2 osd: 18 osds: 18 up (since 9h), 18 in...

Another observation, starting more benchmarks doesnt affect the performance of a single vm that much. but I can see the increased iops under DC->CEPH . There is high overhead in proxmox/vm . I created a new VM with more vcpu, but more importantly these settings: now the VM performance is...

Running the tests in parallell yileds higher total, so it seems that the VM is the most limiting factor here!

Just to compare I setup a lxc with 16vcpu Jobs: 3 (f=3): [_(1),m(2),_(3),m(1),_(1)][100.0%][r=73.2MiB/s,w=7872KiB/s][r=4683,w=492 IOPS][eta 00m:00s] randrw: (groupid=0, jobs=8): err= 0: pid=718: Wed Jan 10 21:17:07 2024 read: IOPS=5626, BW=87.9MiB/s (92.2MB/s)(14.4GiB/167758msec) slat...

overview during testing

changing vcpu doesnt matter much at these low speeds it seems. using sync=0 ( well no sync) yields 800ish write iops @test:~$ fio --name=randrw --rw=randrw --direct=1 --ioengine=libaio --bs=16k --numjobs=4 --rwmixread=75 --size=512MB --runtime=100 --group_reporting randrw: (g=0): rw=randrw...

test with 8 vcpu @test:~$ fio --name=randrw --rw=randrw --direct=1 --ioengine=libaio --bs=16k --numjobs=8 --rwmixread=75 --size=512MB --runtime=100 --sync=1 --group_reporting randrw: (g=0): rw=randrw, bs=(R) 16.0KiB-16.0KiB, (W) 16.0KiB-16.0KiB, (T) 16.0KiB-16.0KiB, ioengine=libaio, iodepth=1...

Nexus 5548 lacp Ill try writeback! and i upped the cores to 8, altought cpu util was 50-75% during testing. nexus 5548UP, see above iperf benchmark, performance is 10Gbit, LACP/vPC

VM settings for the vm that runs fio

mostly differnt dbs in different vms, more io work, its a labb for servers/networking etc. but decent performance is preferred. Ill try and get more local storage for when performance is really needed. with the current setup with 6 sas ssds one each node, that can push (peak iops) 50k/25k iops...

Network benchmark: @pm2:~# iperf -c 172.16.50.10 -i 2 -e ------------------------------------------------------------ Client connecting to 172.16.50.10, TCP port 5001 with pid 89725 (1 flows) Write buffer size: 131072 Byte TOS set to 0x0 (Nagle on) TCP window size: 16.0 KByte (default)...

Running a a labb ( so no real prod usecase, except nice to have) 3 node ceph cluster a 6 hgst sas ssd 200GB, standard setup with 3 replikas. on a 2x10Gbps network, shared network with vms, but there are no vms here yet, so very little other traffic. Ran this fio benchmark in a ubuntu vm, is...