Arvin Schnell wrote:

On Fri, Dec 17, 2010 at 08:40:22PM +0100, Per Jessen wrote:

...

I've just today received a couple of Western Digital 3Tb drives, so I thought it would be worth a status report - initially on 11.3, but I'll be trying out 11.4M5 very shortly:

Platform - very plain, second-hand office desktop IBM M52 with a 2.8GHz Celeron and 2Gb memory. Drive - Western Digital WD30EZRSDTL, 5400rpm, 64Mb cache, 4096 physical sector size. SATA II.

Installation of minimal server went just fine, no problems whatsoever. Note - boot manager lilo, NOT grub.

Good to hear. Where is the partition with /boot located? Entirely below 2 TB?

Yes right at the beginning of the drive, I might try 11.4M5 with a few different partitionings.

...

The partitioning seems a little odd - the first partition starts at 2048s, which is 2048 512-byte sectors, so at 1048576bytes.

That is the default alignment of parted 2.2. It's a feature.

Okay. With 3Tb who cares about 1Mb ... :-) The remaining three partitions all start on a sector# that is a multiple of 8, so that seems fine. I always install with firewall disabled, but I thought sshd should be started automatically - which it wasn't. Moving on to 11.4M5. -- Per Jessen, Zürich (-2.5°C) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Fri, 17 Dec 2010 22:04:03 +0100 Lars Müller <lmuelle@suse.de> wrote:

This thread has all arguments and might help to keep the discussion as short as it's worth. The default was already changed with openSUSE 11.2

The installer UI is still broken: * "open ssh port and enable sshd" * "disable firewall" * install * firewall disabled, but sshd not started (if you first disable the firewall, there is no way to start sshd) No, I'm not going to file a bug, I just accepted that we are going the same crap way as ubuntu. <mumble> not starting sshd, what a cracksmoker's idea </mumble> Now I "open ssh port" in install, then later simply "rpm -e SuSEfirewall2" but it still seems wrong. -- Stefan Seyfried "Any ideas, John?" "Well, surrounding them's out." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Stefan Seyfried wrote:

On Fri, 17 Dec 2010 22:04:03 +0100 Lars Müller <lmuelle@suse.de> wrote:

...

This thread has all arguments and might help to keep the discussion as short as it's worth. The default was already changed with openSUSE 11.2

The installer UI is still broken:

* "open ssh port and enable sshd" * "disable firewall" * install * firewall disabled, but sshd not started

(if you first disable the firewall, there is no way to start sshd)

https://bugzilla.novell.com/show_bug.cgi?id=537980

No, I'm not going to file a bug, I just accepted that we are going the same crap way as ubuntu.

<mumble> not starting sshd, what a cracksmoker's idea </mumble>

One more thing you can tell your grandchildren after talking about the good old times when gcc was still installed by default and autologin disabled ;-)

Now I "open ssh port" in install, then later simply "rpm -e SuSEfirewall2" but it still seems wrong.

# SuSEfirewall2 off # chkconfig sshd on cu Ludwig -- (o_ Ludwig Nussel //\ V_/_ http://www.suse.de/ SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nuernberg) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Stefan Seyfried wrote:

On Mon, 20 Dec 2010 12:04:37 +0100 Ludwig Nussel <ludwig.nussel@suse.de> wrote:

...

One more thing you can tell your grandchildren after talking about the good old times when gcc was still installed by default and autologin disabled ;-)

Actually, gcc is easily selected during installation and autologin is asking for it being disabled by activating network authentication. But SSH is not easily enabled, if you do not want SuSEFirewall ;)

If you install over ssh, ssh will be automatically enabled when the firewall is disabled. -- Per Jessen, Zürich (3.2°C) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Lars Müller wrote:

On Fri, Dec 17, 2010 at 09:08:09PM +0100, Per Jessen wrote: [ 8< ]

...

I always install with firewall disabled, but I thought sshd should be started automatically - which it wasn't.

* Thu Apr 23 2009 lnussel@suse.de - do not enable sshd by default

See also http://lists.opensuse.org/opensuse/2009-11/msg01589.html

This thread has all arguments and might help to keep the discussion as short as it's worth.

Having reread that thread, it somehow doesn't appear to have been much of a community decision. At least it seems to be mostly you arguing for it, Lars. Perhaps it was _your_ decision. Regardless, as our strategy is now different, perhaps I should open a feature request for enabloing sshd by default. (sorry, can't be bothered, it's just getting to be too silly). -- Per Jessen, Zürich (-2.8°C) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Fri, Dec 17, 2010 at 3:08 PM, Per Jessen <per@opensuse.org> wrote:

Arvin Schnell wrote:

...

On Fri, Dec 17, 2010 at 08:40:22PM +0100, Per Jessen wrote:

...

I've just today received a couple of Western Digital 3Tb drives, so I thought it would be worth a status report - initially on 11.3, but I'll be trying out 11.4M5 very shortly:

Platform - very plain, second-hand office desktop IBM M52 with a 2.8GHz Celeron and 2Gb memory. Drive - Western Digital WD30EZRSDTL, 5400rpm, 64Mb cache, 4096 physical sector size. SATA II.

Installation of minimal server went just fine, no problems whatsoever. Note - boot manager lilo, NOT grub.

Good to hear. Where is the partition with /boot located? Entirely below 2 TB?

Yes right at the beginning of the drive, I might try 11.4M5 with a few different partitionings.

...

...

The partitioning seems a little odd - the first partition starts at 2048s, which is 2048 512-byte sectors, so at 1048576bytes.

That is the default alignment of parted 2.2. It's a feature.

Okay. With 3Tb who cares about 1Mb ... :-) The remaining three partitions all start on a sector# that is a multiple of 8, so that seems fine.

A lot new high capacity drives actually have 4 KB physical sectors and 512 byte logical sectors. ie. The smallest internal atomic write they can actually make is 4KB. That's because they have an internal CRC for each physical sector and if any of the data changes, they have to update the CRC. So a single logical sector write will actually trigger a read-modify-write sequence of the physical sector. Therefore it is very important that filesystem pages be aligned with the underlying physical sectors. Therefore parted 2.2 was updated to interrogate the drive for the physical sector details and ensure partitions start on physical sector boundaries. Once the drive is formated the alignment is set, so you don't have to worry about it. Assuming your drive has 4KB physical sectors, you will want to ensure you only partition it on 11.3 or newer. If you did it with 11.2 or older you would see a horrible performance drop due to all the read-modify-write cycles introduced by non-alignment of pages. Greg -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Sat, Dec 18, 2010 at 4:15 AM, Per Jessen <per@opensuse.org> wrote:

Greg Freemyer wrote:

...

...

...

...

The partitioning seems a little odd - the first partition starts at 2048s, which is 2048 512-byte sectors, so at 1048576bytes.

That is the default alignment of parted 2.2. It's a feature.

Okay. With 3Tb who cares about 1Mb ... :-) The remaining three partitions all start on a sector# that is a multiple of 8, so that seems fine.

A lot new high capacity drives actually have 4 KB physical sectors and 512 byte logical sectors.

Yes, as does this one.

...

Therefore parted 2.2 was updated to interrogate the drive for the physical sector details and ensure partitions start on physical sector boundaries.

Once the drive is formated the alignment is set, so you don't have to worry about it.

Assuming your drive has 4KB physical sectors, you will want to ensure you only partition it on 11.3 or newer.  If you did it with 11.2 or older you would see a horrible performance drop due to all the read-modify-write cycles introduced by non-alignment of pages.

Or use a suitable parted?

If and only if you have a current enough kernel, libraries, etc. I think the disk topography patches went into 2.6.32. If so, 11.2 uses 2.6.31 and won't have them. I know the standard 11.2 userspace does not have disk topography support. But you can always enforce the 1MB first partition offset and 4KB alignment on all others manually. Note, SSDs often want 128KB alignment for partitions due to Erase Blocks. (I'm not sure how important that is.) To get that right needs the same drive topography support as your 4KB physical sectors. Greg -- Greg Freemyer Head of EDD Tape Extraction and Processing team Litigation Triage Solutions Specialist http://www.linkedin.com/in/gregfreemyer CNN/TruTV Aired Forensic Imaging Demo -    http://insession.blogs.cnn.com/2010/03/23/how-computer-evidence-gets-retriev... The Norcross Group The Intersection of Evidence & Technology http://www.norcrossgroup.com -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 12/21/2010 01:39 AM, Per Jessen wrote:

Greg Freemyer wrote:

...

On Sat, Dec 18, 2010 at 4:15 AM, Per Jessen <per@opensuse.org> wrote:

...

Greg Freemyer wrote:

...

Assuming your drive has 4KB physical sectors, you will want to ensure you only partition it on 11.3 or newer. If you did it with 11.2 or older you would see a horrible performance drop due to all the read-modify-write cycles introduced by non-alignment of pages.

Or use a suitable parted?

If and only if you have a current enough kernel, libraries, etc.

I think the disk topography patches went into 2.6.32. If so, 11.2 uses 2.6.31 and won't have them. I know the standard 11.2 userspace does not have disk topography support.

Ah, I didn't expect it to be kernel-dependent, I would have thought parted could align the partitions quite well itself.

...

But you can always enforce the 1MB first partition offset and 4KB alignment on all others manually.

That is what I was thinking, and wouldn't a sufficiently up-to-date parted do that for me too?

With care and some manual intervention, it is possible to work with 4K-sector disks on older distros and kernels. I have one such system running openSUSE 11.1 and a 2.6.29 kernel. It mainly runs MythTV, and I have not found the time to upgrade it, and the WAF (wifely admiration factor) would greatly suffer if I broke it. She already thinks Linux is nonsense. As this drive was my second WD Green 1 TB drive, I did not expect any problems, thus I installed, partitioned, and formatted it normally. Only when the performance took a major hit did I discover the nuances of the model number. By manually partitioning and taking care that each one starts at a multiple of 8 of the simulated 512 B sectors, the drive works fine. Larry -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On 12/21/2010 11:55 AM, Stefan Seyfried wrote:

On Tue, 21 Dec 2010 09:01:09 -0600 Larry Finger <Larry.Finger@lwfinger.net> wrote:

...

As this drive was my second WD Green 1 TB drive, I did not expect any problems, thus I installed, partitioned, and formatted it normally. Only when the performance took a major hit did I discover the nuances of the model number.

Now you're scaring me. Which WD Green 1TB drives are 4k and which are old-fashioned?

I have a WD10EADS and it has only one partition on it:

server:~ # parted /dev/sdc print Model: ATA WDC WD10EADS-00L (scsi) Disk /dev/sdc: 1000GB Sector size (logical/physical): 512B/512B Partition Table: msdos

Number Start End Size Type File system Flags 1 32,3kB 1000GB 1000GB primary xfs type=83

I did not notice a major performance problem until now...

The device with 4KB sectors: WDC WD10EARS-00Y5B1. The device with 512B sectors: WDC WD10EADS-00L5B1. Your EADS is OK. If you had the problem, you would notice it. On mine, there would be a 30+ second pause when doing the final linking of modules during a kernel build while the drive caught up with all those split-sector writes. Larry -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

On Tue, 21 Dec 2010 14:05:43 -0500 Greg Freemyer <greg.freemyer@gmail.com> wrote:

On Tue, Dec 21, 2010 at 12:55 PM, Stefan Seyfried

...

server:~ # parted /dev/sdc print Model: ATA WDC WD10EADS-00L (scsi) Disk /dev/sdc: 1000GB Sector size (logical/physical): 512B/512B

Assuming the drive is telling the truth, its the traditional style 512B physical sector.

The new drives should report 512B/4096B I believe.

Ok. Together with Larry's hints, this makes me confident that I'm not mistreating it badly ;) With some XFS tuning, it is even possible to "rm -rf" a kernel tree in less than 10 minutes ;-) server:~ # mount | grep sdc /dev/sdc1 on /space2 type xfs (rw,noatime,logbufs=8,logbsize=262144)

But those green drives are nasty regardless. They tend to have very short idle periods before unloading the head. I've even seen error reports in windows where they cause applications to malfunction. Truly bad drives IMHO.

not mine. It's about a year old, I think, and it has reasonable values: server:~ # smartctl -A /dev/sdc smartctl 5.39.1 2010-01-28 r3054 [x86_64-unknown-linux-gnu] (openSUSE RPM) Copyright (C) 2002-10 by Bruce Allen, http://smartmontools.sourceforge.net === START OF READ SMART DATA SECTION === SMART Attributes Data Structure revision number: 16 Vendor Specific SMART Attributes with Thresholds: ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE 1 Raw_Read_Error_Rate 0x002f 200 200 051 Pre-fail Always - 0 3 Spin_Up_Time 0x0027 158 157 021 Pre-fail Always - 7058 4 Start_Stop_Count 0x0032 099 099 000 Old_age Always - 1693 5 Reallocated_Sector_Ct 0x0033 200 200 140 Pre-fail Always - 0 7 Seek_Error_Rate 0x002e 100 253 000 Old_age Always - 0 9 Power_On_Hours 0x0032 081 081 000 Old_age Always - 14154 10 Spin_Retry_Count 0x0032 100 100 000 Old_age Always - 0 11 Calibration_Retry_Count 0x0032 100 253 000 Old_age Always - 0 12 Power_Cycle_Count 0x0032 100 100 000 Old_age Always - 61 192 Power-Off_Retract_Count 0x0032 200 200 000 Old_age Always - 25 193 Load_Cycle_Count 0x0032 200 200 000 Old_age Always - 1689 194 Temperature_Celsius 0x0022 128 100 000 Old_age Always - 22 196 Reallocated_Event_Count 0x0032 200 200 000 Old_age Always - 0 197 Current_Pending_Sector 0x0032 200 200 000 Old_age Always - 0 198 Offline_Uncorrectable 0x0030 200 200 000 Old_age Offline - 0 199 UDMA_CRC_Error_Count 0x0032 200 200 000 Old_age Always - 0 200 Multi_Zone_Error_Rate 0x0008 200 200 000 Old_age Offline - 0 As you can see, the Load_Cycle_Count is about the same as the Start_Stop_Count. The only thing I miss on this drive is that it is not spinning down on its own - no matter what values you put into hdparm -S, it ignores them IIRC. It also does not auto-spindown with an aggressively low hdparm -B value. So I wrote a little C program which checks /proc/diskstats and if it does not change for this drive for > X seconds, it runs "hdparm -y" on the drive. Works well for me. But I have to admit that this drive is in my server box as a storage drive only - I have a 2.5" drive for the system and primary data partition, so it usually only spins one hour per day, recording my wife's daily soap just in case she's missing it ;) All the "heavy lifting" - albeit in smaller quantities - is done by the 2.5" drive. IMHO, the 1TB green drive delivers top performance for little juice. But then, that's just my experience.

I think the smart param that tracks the issue is Load_Cycle_Count. Most drives are only rated for 50,000 or so of those per lifetime.

They were rated 350000 10 years ago - I noticed that on an old 40GB 2.5" toshiba drive which did loud "CLONK" every 3 seconds out of the cabinet. I wondered how a device this small could generate a sound that loud. While investigating who had switched such a feature on, I found out that it was the "intelligent" auto-parking-feature with APM level 128 (which was the default on this drive), that decided it was a good idea to park the head after 3 seconds of non-activity - with an ext3 commit interval of 5 seconds, that was deadly. OTOH, with disabled auto-park this drive is still running fine today, with 5 million load cycles while being specified for 350000. I kept away from toshiba drives ever since ;) (Note that I actually investigated the "intelligent" power management functions including auto-spindown with APM level < 128 later and found, that modern drives are doing pretty good which actually lead me to not messing around with hdparm -S anymore and just letting the drive decide when it's time to fall asleep). Todays drives still show high (means: good) smart values with 500000 load cycles, so I guess they are now rated much higher. And even windows does not prevent the automatic head parking on those drives, so I guess tha

The green drives without special handling under linux will use those up in a year.

Now that you mention it, that's what I have in boot.local: GB320=/dev/disk/by-id/scsi-SATA_FUJITSU_MHZ2320_K60LT912RPY6 GB500=/dev/disk/by-id/scsi-SATA_MAXTOR_STM35006_9QG3JP5T GB1000=/dev/disk/by-id/scsi-SATA_WDC_WD10EADS-00_WD-WCAU46193879 hdparm -B 192 $GB320 # load cycle count reduction hdparm -S 240 $GB500 hdparm -S 240 $GB1000 (the FUJITSU drive was recently replaced by a WDC WD7500BPVT, so I need to adopt that script - and indeed, that 750G 2.5" drive has 268000 cycles already, so I'll have to adjust that :), the WD10EADS ignores the hdparm -S setting, so I could just remove that)

Personally I just avoid the green drives like the plague, but since you have one you should look into Load_Cycle_Count and if its too high get your drive added to "storage.fixup" I don't know if that's in kernel or userspace, but it is a config file, etc. that is used to tell the linux kernel that the drive is not well behaved and to take special action to keep it from unloading the heads excessively.

It's a userspace tool that usually runs during boot and resume.

Tejun Heo is the maintainer for storage.fixup I believe.

I know, I helped him packaging it initially ;-) (Totally offtopic now ;-) Thanks for all the hints, Stefan -- Stefan Seyfried "Any ideas, John?" "Well, surrounding them's out." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org

Per Jessen wrote:

Per Jessen wrote:

...

Arvin Schnell wrote:

...

Good to hear. Where is the partition with /boot located? Entirely below 2 TB?

Yes right at the beginning of the drive, I might try 11.4M5 with a few different partitionings.

No problems in 11.4M5 either - I moved the boot partition up to around 1.6Tb, no problems.

I then tried booting from partition #4 at around 1.6Tb, which I think produced some warning, and later on didn't work either. Initially I also had to run lilo with '-P ignore' due to the problem about some CHS/LBA mismatch. This problem later disappeared after I manually reconfigured the partitioning with parted. -- Per Jessen, Zürich (-2.2°C) -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-factory+help@opensuse.org