Solaris sux

[Jakes Louw]

Sorry, this isn't a question or a query, just an observation born out of frustration...
I'm certified on HP-UX and Solaris, and I'm currently working as a consultant for Sun, and I've come to the conclusion that Sun's products are awesome hardware-wise, but those propeller-heads at the Labs that think up disk management methods are not well.
Sun has no single, consistent disk management method, and nothing talks to each other. To find out what is going on, you need to run Explorer and dig thru 500 pages of garbage before you can understand the volume layout. Having worked extensively on top range HP-UX servers, clustered and not, I've got to say there is only ONE Solaris facility that HP-UX should have: the "GLOBAL" mount option for shared cluster disks. It beats VGIMPORT hands-down any day.
Comments?

[F Verschuren · Accepted solution]

You are right, but veritas "sux" is as bad on hp as on sun.....

[Jakes Louw]

SolVM is just as bad. And then you have SCDIDADM for clustered disks, and and and and

[Arunvijai_4]

May be you are right. I don't think "vgimport" is so bad ??

-Arun

[Jakes Louw]

VGIMPORT isn't so bad? Try mixing your VGIDs within an MC/SG Cluster, then having to change the VGID before you can do the IMPORT by modifying map files.....

[Chan 007]

Hi,

Why you need to modify, you -s when doing the vgexport, there is no need to modift your map file as it does your VGCHGID, you feel vgimport is better

Chan

[Florian Heigl (new acc)]

We once had some Ex-consultant working in our team, who had done some solaris work, and kept babbling that 'veritas is a next-generation volume manager'. One day I took VXFS manuals and started playing with my tru64 cluster.

Honestly veritas is a step backward for anyone who no longer is using disk slices (aix, hp-ux), but at least after that I understood why the sun/tru64 users did like it - it was an improment for them.
for anyone else it's just a few additional layers of garbage. :)


Random comments:
- ZFS will hopefully get us rid of both solstice and veritas in the long term.
- SGI CXFS and FreeBSD's GEOM is the only 'next-generation' thing at all I could think of.
- Linux is even more mess now, with sw-raid, lvm, lvm2, evms and everything layered above and below everything else.

[Geoff Wild]

As far as the "glbal" mount - there is - in HPUX 11iV2 and ServiceGuard 11.17 - it is called the Clustered File System.

Some notes on it:

Allows multiple hosts simultaneous read/write access to file systems
Adds an additional disk volume configuration option for an applicationâ s data within a SG cluster but retains previous options
Within a given SG/CFS Cluster different applications could use any combination of the following mechanisms:
Local raw volumes that failover
Local file systems that failover
Shared raw volumes (e.g., Oracle RAC)
Shared file systems (CFS)
LVM and VxVM/CVM volumes can coexist in the same cluster but CFS file systems must use CVM
Not for root file system (/, /etc, /sbin, /usr, /var, etc.)

Rgds...Geoff

[Jakes Louw]

Cool, Geoff, I haven't been able to work on anything after HP-X 11.11, so that's good news. I've given most of you 10 points, those of you who I didn't it's because U are off the mark, but close. I suppose I can close this sucker.

[Zinky]

Peace Jakes,

That "ONE Solaris" facility is actually Veritas (Veritas Cluster) and you're right just like HP-UX and other Unices, Solaris offers "choices" for volume management -- with Veritas still being considered the best until probably when ZFS comes out.

VxVM's diskgroups do not rely on phsyical volume paths that can change accross reboots or "mysteries of the SAN" and it is in this area that it beats LVM's vgimport -- where you must need to keep a track and backups of your VGs, VGIDS and lvols.

Veritas VxVM, VxFS Filesystem (which is already OEM'd to HP-UX as JFS/OnlineJFS) and Cluster are the reasons why Tru64's features - filesystem, volume management and clustering technologies , were not incorporated to with HP-UX as originally planned. Why? -- Because the Veritas technologies are already available and already works and would require less integration as VxVM and VxFS are already in place (since 11.11 '2002 I think). Also, since VxVM/VxFS are already widely popular on other UNICES - data centers and enterprises can have a common interface to storage management accross practically all operating environments.

I started out with Linux LVM and did fing it difficult adjusting to other Volume Managers on Tru64, NCR, HP-UX, AIX and Solaris. If you know the uderlying technologies of storage organization -- i.e. LUNS, physical disks, RAID techniques, etc. -- it should be easy to adjust in other volume manager environments.

[Steven E. Protter]

Shalome Jakes,

I agree on the concept of LVM being a good, unified management system.

The fact that LVM requires you to vgimport to get access to the volume group prior to mounting is a small price to pay for the benefits that lvm provides.

Linux now provides GFS, which allows cluster clients to mount with a simple mount_gfs command. This is similar to what you wish and Linux has LVM ported quite nicely.

SEP

[dirk dierickx]

both hpux and sun have crappy volume manager support by default. Solstice is hardly a volume manager in my book.

zfs looks very promising though, judging by the demos i saw online.

[Jakes Louw]

This was fun, but the thread is now closed......will the best UNIX flavour please come forward? ;->