Write is a good thing for NetApp.
We turn random writes into sequential.
You should know what FVBN and VVBN is.
NetApp writes data down to disk in the most efficient way.
NetApp benefits massively with writes.
NetApp tends to write a little bit more.
NetApp writes to system memory, we do not write to NVRAM.
Perfstat: Total NetIn etcetera - can use to calculate average IO size.
Ignore cache age (multiple caches in modern NetApp systems.)
CP_Ty column in perfstat is important to understand.
T= Timed ...
Google "sysstat manual" for explanations. Like this link:
https://library.netapp.com/ecmdocs/ECMP1511538/html/man1/na_sysstat.1.html
CP time 100% 100% 100% and very little break, I would be getting concerned when is it going to write the CP. Worried might get to the BB (back-to-back CP) scenario (which is very bad.)
HDD util (in perfstat), skip. It is the most utilized disk in the entire system, not useful information.
Protocol: net/in net/out...
SnapMirror is not listed in OTHER, only net-in and net-out can you see SnapMirror.
If you have a CPU pegged (with multiple core view) you know there's some domain limitation.
Kahu is single threaded. Kahu can never go above 99%.
statit -b # to begin
statit -e # to end
Amazing useful.
Too many IOPS on a single disk (from the statit) is a sign of trouble (i.e. 200 IOPS on a 10k disk.)
RAID statistics. Stripes. Lots of full stripes is good news (like over 50% very good, over 66% exceptional - like a new system.)
We turn random writes into sequential.
You should know what FVBN and VVBN is.
NetApp writes data down to disk in the most efficient way.
NetApp benefits massively with writes.
NetApp tends to write a little bit more.
NetApp writes to system memory, we do not write to NVRAM.
Perfstat: Total NetIn etcetera - can use to calculate average IO size.
Ignore cache age (multiple caches in modern NetApp systems.)
CP_Ty column in perfstat is important to understand.
T= Timed ...
Google "sysstat manual" for explanations. Like this link:
https://library.netapp.com/ecmdocs/ECMP1511538/html/man1/na_sysstat.1.html
CP time 100% 100% 100% and very little break, I would be getting concerned when is it going to write the CP. Worried might get to the BB (back-to-back CP) scenario (which is very bad.)
HDD util (in perfstat), skip. It is the most utilized disk in the entire system, not useful information.
Protocol: net/in net/out...
SnapMirror is not listed in OTHER, only net-in and net-out can you see SnapMirror.
If you have a CPU pegged (with multiple core view) you know there's some domain limitation.
Kahu is single threaded. Kahu can never go above 99%.
statit -b # to begin
statit -e # to end
Amazing useful.
Too many IOPS on a single disk (from the statit) is a sign of trouble (i.e. 200 IOPS on a 10k disk.)
RAID statistics. Stripes. Lots of full stripes is good news (like over 50% very good, over 66% exceptional - like a new system.)
Comments
Post a Comment