nvme-cli: nvmf-autoconnect: udev-rule: add a file for new arrays
Subsystems (3)
PATCH 1/1 nvme-cli: nvmf-autoconnect: udev-rule: add a file for new arrays
One file per vendor, or device, is a bit excessive for two-four rules.
If possible, select round-robin (>=5.1), or queue-depth (>=6.11). round-robin is a basic selector, and only works well under ideal conditions.
A nvme benchmark, round-robin vs queue-depth, shows how bad it is:
https://marc.info/?l=linux-kernel&m=171931850925572
https://marc.info/?l=linux-kernel&m=171931852025575
https://github.com/johnmeneghini/iopolicy/?tab=readme-ov-file#sample-data
https://people.redhat.com/jmeneghi/ALPSS_2023/NVMe_QD_Multipathing.pdf
And it is unnecessary to set ctrl_loss_tmo to -1 for NVMe/TCP devices, because it is the default.
---
Maybe these rules should be merged into this new file. ??? 71-nvmf-hpe.rules.in
71-nvmf-netapp.rules.in
71-nvmf-vastdata.rules.in
--- .../80-nvmf-storage_arrays.rules.in | 33 +++++++++++++++++++ 1 file changed, 33 insertions(+) create mode 100644 nvmf-autoconnect/udev-rules/80-nvmf-storage_arrays.rules.in
nvmf-autoconnect/udev-rules/80-nvmf-storage_arrays.rules.in new file +33 -0
@@ -0,0 +1,33 @@ +##### Storage arrays + +#### Set iopolicy for NVMe-oF +### iopolicy: numa, round-robin (>=5.1), or queue-depth (>=6.11) + +## Dell EMC +# PowerMax +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="round-robin", ATTR{model}=="EMC PowerMax" +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="queue-depth", ATTR{model}=="EMC PowerMax" +# PowerStore +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="round-robin", ATTR{model}=="dellemc-powerstore" +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="queue-depth", ATTR{model}=="dellemc-powerstore" + +## Huawei +# OceanStor +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="round-robin", ATTR{model}=="Huawei-XSG1" +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="queue-depth", ATTR{model}=="Huawei-XSG1" + +## IBM +# FlashSystem (RamSan) +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="round-robin", ATTR{model}=="FlashSystem" +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="queue-depth", ATTR{model}=="FlashSystem" +# FlashSystem (Storwize/SVC) +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="round-robin", ATTR{model}=="IBM*214" +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="queue-depth", ATTR{model}=="IBM*214" + +## Pure +# FlashArray +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="round-robin", ATTR{model}=="Pure Storage FlashArray" +ACTION=="add|change", SUBSYSTEM=="nvme-subsystem", ATTR{subsystype}=="nvm", ATTR{iopolicy}="queue-depth", ATTR{model}=="Pure Storage FlashArray" + + +##### EOF
Signature · 1 line
2.50.1
Quoted 2 lines · Xose Vazquez Perez
And it is unnecessary to set
ctrl_loss_tmoto -1 for NVMe/TCP devices, because it is the default.
Totally wrong. I misread the code:
$ grep NVMF_DEF_CTRL_LOSS_TMO drivers/nvme/host/fabrics.c
int ctrl_loss_tmo = NVMF_DEF_CTRL_LOSS_TMO, key_id;$ grep ^#define drivers/nvme/host/fabrics.h #define _NVME_FABRICS_H 1 #define NVMF_MIN_QUEUE_SIZE 16 #define NVMF_MAX_QUEUE_SIZE 1024 #define NVMF_DEF_QUEUE_SIZE 128 #define NVMF_DEF_RECONNECT_DELAY 10 #define NVMF_DEF_CTRL_LOSS_TMO 600 <--- #define NVMF_DEF_FAIL_FAST_TMO -1
It's 10 minutes.