nndocs:iscsi
Differences
This shows you the differences between two versions of the page.
Both sides previous revisionPrevious revisionNext revision | Previous revision | ||
nndocs:iscsi [2024/02/28 14:55] – [IQNs] clarify legacy IQN naming naptastic | nndocs:iscsi [2025/01/03 02:32] (current) – [Initiator] naptastic | ||
---|---|---|---|
Line 1: | Line 1: | ||
- | TODO: Make all the initiator and target names match please | ||
- | |||
TODO: Explain why I don't have iSCSI boot working yet, please | TODO: Explain why I don't have iSCSI boot working yet, please | ||
- | =====IQNs===== | ||
- | We do IQNs by the book, but as simply as possible. A couple of examples: | ||
- | |||
- | iqn.2014-08.rocks.narf.sadness | ||
- | iqn.2004-12.com.naptastic.quirk | ||
- | |||
- | IQNs can also have :more.stuff on the end. I can't come up with a use case, since hostnames are already fully-qualified and unique. | ||
- | |||
- | There are still some :01 suffixes lurking around the network. Fix them or don't; just don't make new ones. | ||
- | |||
- | Ideally, / | ||
=====Target===== | =====Target===== | ||
Line 28: | Line 15: | ||
> portals/ create 172.19.12.1 3260 | > portals/ create 172.19.12.1 3260 | ||
> portals/ | > portals/ | ||
- | |||
- | TODO: enable_iser doesn' | ||
=====Initiator===== | =====Initiator===== | ||
Line 37: | Line 22: | ||
===Login. Must use a node record id found by the discovery=== | ===Login. Must use a node record id found by the discovery=== | ||
- | iscsiadm --mode node --targetname iqn.2001-05.com.doe: | + | iscsiadm --mode node --targetname iqn.2014-08.rocks.narf.southpark |
[[https:// | [[https:// | ||
Line 44: | Line 29: | ||
# iscsiadm -m session | # iscsiadm -m session | ||
- | tcp: [1] 172.21.0.80: | + | tcp: [1] 172.20.0.80: |
The session number (sid) is the 1 in brackets. | The session number (sid) is the 1 in brackets. | ||
===Use iSER instead of TCP (initiator side)=== | ===Use iSER instead of TCP (initiator side)=== | ||
- | If performance is poor, the initiator may have logged in using TCP instead of iSER. In my testing, to a ramdisk over IB, TCP got ~520 MiB/s, where iSER got 1.9 GiB/s. | + | If performance is poor, the initiator may have logged in using TCP instead of iSER. In my testing, to a ramdisk over IB, TCP got ~520 MiB/s, where with iSER I have gotten 3.2 GB/s. |
You know you got TCP if dmesg describes your LUN like this: | You know you got TCP if dmesg describes your LUN like this: | ||
Line 64: | Line 49: | ||
[123.45678] scsi host1: iSCSI Initiator over iSER | [123.45678] scsi host1: iSCSI Initiator over iSER | ||
[123.45678] scsi 1:0:0:0: Direct-Access | [123.45678] scsi 1:0:0:0: Direct-Access | ||
- | |||
- | FDR InfiniBand should be able to move over 6 GB/s of traffic. 1.9 GiB/s is nothing to be sad about. | ||
===Rescan after the target adds a LUN=== | ===Rescan after the target adds a LUN=== | ||
# iscsiadm -m session --sid=1 --rescan | # iscsiadm -m session --sid=1 --rescan | ||
- | Rescanning session [sid: 1, target: iqn.2003-01.org.linux-iscsi.happy.x8664: | + | Rescanning session [sid: 1, target: iqn.2014-08.rocks.narf.southpark, portal: 172.20.0.80,3260] |
===Log out when you're done (very, very carefully)=== | ===Log out when you're done (very, very carefully)=== |
nndocs/iscsi.1709132128.txt.gz · Last modified: 2024/02/28 14:55 by naptastic