Category Archives: Server administration

Sysadmins Don’t buy 3TB Seagate Hard Disks! – 3TB hard disk huge disk failure reports

Dont-buy-Seagate-3-TB-HDDs-or-you-will-have-big-problems-with-server-data-and-downtimes
Seagate Hard disks of 3 Terabyte volume used to be very attractive HDDs (storage) option when they were launched on the market.
Many people and sysadmins have already bought such ones and some sysadmins and company customers that already choose Seagate are badly suffering because of that and thus it is good to warn others to stay away from 3TB SeaGate Hard Disks.

Backblaze (Online Backup) company is one of the most severely affected companies that made the choise to use Seagate as a storage devices on their Cloud inter-connected servers. They used 41 213 hard disks in their computing Data Center as of 31 December 2014.

In their disk arrays they have used Western Digital  (now part of Western DIgital) and of course pitily Seagate.

The problematic hard disks that they faced issues with are Seagate Barracuda 7200.14 3TB sized is the hard disks with most failures within Backblaze for whole 2014 about 40% percents!! of all 3TB hard disks the company had break up, died or had to be replaced because of I/O disk failures and bad-sectors.

It is not exactly clear what is the reason for such a high failures but Seagate were leaders in failures followed by Western Digital and HGTX (the ex-Hitachi).

 


Just for a comparison Backblaze reports that 4 Tetabyte hard disks which they bought last year had failures, very rarely and in general the company is quite happy with Seagate / WD disks of 4 TB volume.

Seagate Barracuda 7200.14 3TB disk drives mounted on their servers are the one who had most hardware issues and the company recommends anyone willing to buy a new HDD to stay away from this volume.

Western Digitals 3TB HDDs had 10% of failure rate, HGTS  had only 2.6% and Seagate exact failed HDDs were approximately 43.1% with a HDD failure!!

No severe hardware HDD failures are reported with 4 TB hdds.
4TB Seagate HDDs gave 5% of defects, followed by WD with 3-4% and HGTS with only 1.4%.

Statistics clearly shows it if you want to buy a big storage for your big data / Web / FTP / Dropbox (Cloud) hosting Company as of time of writting 26.01.2015 it is better equip your Big Storage Array racks with HGTS branded hard drives.

 

Download this Article in PDF: Download PDF

How to utilize better processor work on Linux server by disable CPU scaling – Make Linux server CPU work on full speed

how-to-utilize-better-linux-cpu-by-disabling-cpu-scaling-make-linux-server-cpu-work-on-full-speed
I’ve learned a very useful system administration tip on how to utilize processor work on a Linux server. By default in newer Linux kernel releases, there is ondemand CPU function enabled that makes use  of multi-CPU processors only in case if the CPU is needed. On busy servers this on-demand practice is very bad practice, because often when ondemand is enabled CPU is working on a lower CPU rate than maximum (to save power)  and used only when necessery and this is time consuming and affects performance.
To check whether ondemand “kernel feature” is making your CPU work by default on its maximum capacity:

uname -a;
Linux pcfreak 2.6.32-5-amd64 #1 SMP Sun Sep 23 11:00:33 UTC 2012 x86_64 GNU/Linux

 ps ax| grep kondemand
182 ?        S      0:00 [kondemand/0]
183 ?        S      0:00 [kondemand/1]
184 ?        S      0:00 [kondemand/2]
185 ?        S      0:00 [kondemand/3]
186 ?        S      0:00 [kondemand/4]

This processes are controlled by kernel and control ondemand for each server CPU.

grep -E ‘^model name|^cpu MHz’ /proc/cpuinfo
model name      : Intel(R) Core(TM)2 CPU          6600  @ 2.40GHz
cpu MHz         : 1596.000
model name      : Intel(R) Core(TM)2 CPU          6600  @ 2.40GHz
cpu MHz         : 1596.000

As you can see even though the CPU has capacity to run 2.40Ghz only 1.6Ghz are used meaning you’re wasing computing speed and even if used it is slower (not to mention that for dedicated servers saving power is not a priority of the sysadmin).

If ondemand is enabled for CPUs usually this
To disable CPU ondemand function use following one liner:

for CPUFREQ in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do [ -f $CPUFREQ ] || continue; echo -n performance > $CPUFREQ; done

To make the ondemand disabled permanent, quickest way is to add it to /etc/rc.local right before the exit 0

On CentOS Linux to stop ondemand support:

service cpuspeed stop

On RHEL and Oracle Unbreakable Linux:

lsmod | grep ondemand
cpufreq_ondemand        8764  0
freq_table              3751  2 cpufreq_ondemand,acpi_cpufreq
rmmod cpufreq_ondemand acpi_cpufreq freq_table

Any other services that control CPU stepping like ‘cpuspeed’, ‘cpufreqd’, ‘powerd’ should be disabled.

Download this Article in PDF: Download PDF