[lustre-discuss] Lustre File Mirroring Questions

[Brian Lambrigger]

Hello Lustre-discuss,

While testing file mirroring from the FLR initiative, I created the
following setup:

1) Two nodes with an OST each. NodeA has MGS, MDT, and OST0. NodeB has OST1
2) Created two pools on the DFS, with OST0 in pool0, OST1 in pool 1.
3) Created a mirrored layout for the DFS root directory, so that child
files will inherit the layout.
4) Created a file in the directory which inherits the mirroring.


That worked well, but I had some questions as to the capabilities of the
mirrored layout:

1) With the file synced on both OSTs, I downed OST0. On OST1, I was able to
read its replica, but any writes would hang as it attempted to contact
OST0. Is this expected behavior, or do I need to change parameters for the
devices with tunefs in order to have writes failover to OST1?
2) It seems that resync must sync the entire file upon modification, even
if its striped into smaller blocks and only one of those stripes is
modified. Is that correct?


Here's my current OST0000 setup as reported by tunefs:

Read previous values:
Target:     flrtest-OST0000 Index:      0
Lustre FS:  flrtest Mount type: zfs
Flags:      0x2
              (OST )
Persistent mount opts:
Parameters: mgsnode=192.168.7.60 at tcp1 failover.mode=failout


Permanent disk data:

Target:     flrtest-OST0000
Index:      0
Lustre FS:  flrtest Mount type: zfs
Flags:      0x2
              (OST )
Persistent mount opts:
Parameters: mgsnode=192.168.7.60 at tcp1 failover.mode=failout


As a newcomer to Lustre, thanks so much for your time and expertise.

-- 

<http://opendrives.com/>

*Brian Lambrigger* | Platform Developer

10602 Virginia Avenue | Culver City, CA 90232

O: (310) 659-8999 | M: (xxx) xxx-xxxx
<http://opendrives.com/> <http://linkedin.com/company/3184481>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20220510/fb54ede9/attachment.html>