Slow read performance on iscsi SAN

[Testani]

Hi all,
we are going to build a production 3 node cluster use an ISCSI 10gb SAN Lenovo DE.

We are at the beginning and we have configured the first node using two 10gb network cards and a single LUN on the SAN.I attach the multipath configuration, the paths that are active and the performance I obtain.Anyone in my situation or can give me some advice?All mtu are set to 1500, changing to 9000 has no effect.Thank you

here my multipath.conf:


blacklist {


wwid .*


}





blacklist_exceptions {


wwid "36d039ea000bc680e000000ff66fd6ed2"


}





multipaths {


multipath {


wwid "36d039ea000bc680e000000ff66fd6ed2"


alias mpath0


}


}





defaults {


polling_interval 2


path_selector "round-robin 0"


path_grouping_policy multibus


uid_attribute ID_SERIAL


rr_min_io 100


failback immediate


no_path_retry queue


user_friendly_names yes


}








devices {


device {


vendor "LENOVO"


product "DE_Series"


product_blacklist "Universal Xport"


path_grouping_policy "group_by_prio"


path_checker "rdac"


features "2 pg_init_retries 50"


hardware_handler "1 rdac"


prio "rdac"


failback immediate


rr_weight "uniform"


no_path_retry 30


retain_attached_hw_handler yes


detect_prio yes


}


}

 

[bbgeek17]

Hi @Testani , performance troubleshooting and optimization is always multifaceted. There isn't usually a one knob that "fixes" bad performance.

The first step is always to establish a baseline directly on the hypervisor. Then you move on to the VM.

We've written a number of performance oriented articles, you should review them and see what may apply to you:
https://kb.blockbridge.com/technote/proxmox-tuning-low-latency-storage
https://kb.blockbridge.com/technote/proxmox-optimizing-windows-server/part-1.html

Good luck

---

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox

 

[Testani]

thanks, we can't reach more thank 4MB/s in read, i think something big is wrong

 

[Testani]

After several debugging attempts we found the problem in proxmox's iscsi initiator. Using the same bridge with a virtualized windows machine and initializing the iscsi from windows we get 10gb of throughput. By connecting a physical Windows machine the iscsi goes to 10gb, by connecting an ESX the same. The problem lies in the initiator used by proxmox, how can I debug it further?