c0t0d0s0.org

As this article announces events in Germany, this entry will continue in german language. Sorry.

Am 23. Mai wird Hartmut Streppel in der Geschäftsstelle Hamburg einen Vortrag im Rahmen der Oracle Business Breakfasts halten. Thema ist hier "Oracle Solaris Cluster entmystifiziert und virtualisiert". Hartmut ist in Deutschland der Experte zum Thema Oracle Solaris Cluster. Anmelden könnt Ihr euch unter diesem Link.

Einen Tag später - am 24. Mai - gibt es in Potsdam ein weiteres Event im Rahmen der Business Breakfasts. Thema ist hier "ZFS und Datenbanken". Vortragen werden hier mein Kollege Dirk Nitschke und ich. Auch hier ist die Anmeldeseite schon verfügbar. Eine Registrierung ist auf dieser Seite. möglich.

Obwohl offensichtlicherweise nicht in Deutschland, möchte ich bei dieser Gelegenheit möchte ich zudem auch noch auf den "Oracle Solaris Day 2013 Vienna" am 23. Mai hinweisen, auf dem auch eine Reihe von ausgewiesenen Experten sowie meine Wenigkeit vortragen werden. Die Agenda auf der Registrierungsseite ist nicht ganz richtig, da ich später rekrutiert wurde Den Vortrag "What's new in Solaris 11.1?" wird von mir gehalten.

It's just a little tip from the field: A customer has suddenly really bad performance on one of his servers. he finds a lot of error messages like this in the output of dmesg:

Apr 21 05:22:11 server1 scsi: [ID 243001 kern.info] /scsi_vhci (scsi_vhci0):
Apr 21 05:22:11 server1 /scsi_vhci/ssd@g6001438005de946defa2000000020010 (ssd38): Command failed to complete (3) on path fp9/ssd@w50001fe15023ef59,a

On May 23rd an Oracle Solaris event will take place in Vienna. You will find further information about this event in the Oracle Systems Blog Austria: "Solaris Technical Day on the 23rd May in Vienna".

Interesting whitepaper about the compliance to PCI DSS (Payment Card Industry Data Security Standard) with Solaris 11: "Oracle Solaris 11 and PCI DSS"

One of the lesser known features of useradd command is the ability to set defaults for the creation of users, nevertheless is featured prominently on the man page.

A nice explanation how to use OpenLDAP as a name service directory and enabling the administration of the content of the directory (in order to make useradd -S ldap work) in Solaris 11: "Getting Started with OpenLDAP"

Update: "Adding Users with OpenLDAP" gives some additional information in regard of adding users.

SPARC T5 and M5 had also a large impact on the work of the solaris development. Steven Sistare just started a series of article explaining the steps how to scale Solaris to a system with 1536 CPUs or 32 TB of main memory. Two blog entries found their way into his blog so far:

• Massive Solaris Scalability for the T5-8 and M5-32, Part 1 - an introduction
• Massive Solaris Scalability for the T5-8 and M5-32, Part 2 - focused on virtual memory and resource limits.

I've got quite a number of tweets and mails with the question "But is zfs scrub" not something like fsck. And the answer is "Well ... no".

• It's zpool scrub, not zfs scrub. So it can't be a fsck. Perhaps it's a pock, but not fsck
  (Okay, that's a really lame argument. )
• Scrub is not part of the ZFS Posix layer, the entity that let you access a zpool with the semantics of a filesystem. It's done much deeper in ZFS and while it's following a tree there, this tree has nothing to do with your filesystem structure.
  (Ah, not much better reasoning.)
• You don't run zfs scrub on a unimportable pool, you run run zpool clear -F, which starts the txg rollback to make the pool importable again.
  (Don't know if this one is more convincing)
• A scrub works as well on a ZFS Emulated Volume.It can't do a filesystem check on it. Obviously, as Solaris has no idea for example how to check a ext2 file system that you are writing via iSCSI into a zvol. However the scrub has still to work.
  (Perhaps a little better.)
• I've got one tweet saying that it finds silent corruption in the data, thus it's a filesystem check. What I find interesting about this, is the point that this definition would exclude other fsck from being fsck, as they just check the validity of the metadata and not the validity of the data. zpool scrub is a data validity checker, not a filesystem checker. It checks if you are reading the data from a location that you have once written to this location. Whatever this data is. As far as i understand the source, it even doesn't understand the concept of "filesystems" at all.
• What is zpool scrub? It is not much more than reading everything. However the repair mechanism is the same as the one when you read a block and the checksum on-disk doesn't match the checksum computed. The scrub has no own repair code. In the case an error is detected , for example for RAID1 the known good copy is written in the place of the bad copy. Whereas the repair while normal reading is more or less accidental (you've stumbled over an error and reading at one location doesn't mean that other locations are correct, you have to check all locations to be sure). Scrub is the forced search for such incorrectnesses on all copies of a block stored in a pool. With a scrub you ensure that all redundancies are correct. At end scrub checks if the data on-disk is still the data you have written once by other means (ZPL or ZVOL) on it. It simply doesn't care about the structure, as it doesn't know about the structure. It's like memory scrubbing or RAID scrubbing. When your trigger a scrubbing in your RAID array the array gives nothing what's on it.

So, no zfs scrub is no fsck, it has a different job. It might look like a fsck, but it isn't. Forget everything how you expect filesystems internals from your pre-ZFS knowledge. I can give just one important advice in regard of understanding ZFS. An advice that was already used on the first presentations about ZFS: Free your mind!

Interesting annotation about the fuss surrounding the CFS in Linux:

I find it interesting and slightly sad, given how low level a topic this really is, how much is being written about the new CFS scheduler being introduced into Linux. The sad part is how much flamage is flying around as a result of this from people not in the slightest bit involved in the desgin and development - this sadly is the ugly side of many open source groups.

Paul Murphy discuss in his blog the ranking of the different operating environments: OS ranking - confusion to the nth. There are some remarkable quotes.

The problem is simply this: evaluate any Windows OS against Linux, BSD (including MacOS X), or Solaris on any combination of cost, functionality, and reliability and you end up predicting a zero Microsoft market share.

and

So what's the bottom line? There are two of them: first any ranking of multiple OSes is going to have significant subjective components, and secondly that I think it's fair to say that SUse 7.1 was at the top of its class but that Solaris 10 holds that position today.