Re: Why could different data in a table be processed with different performance? - Mailing list pgsql-performance

From Vladimir Ryabtsev
Subject Re: Why could different data in a table be processed with different performance?
Date
Msg-id CAMqTPqmsxoDVW_hvytRbuAJAZoW88ks=o9fd1qK70=2SjR9CLQ@mail.gmail.com
Whole thread Raw
In response to Re: Why could different data in a table be processed with differentperformance?  (Gasper Zejn <zejn@owca.info>)
List pgsql-performance
> 1) Which file system are you using?
From Linux's view it's ext4. Real vmdx file on Hyper-V is stored on NTFS, as far as I know.

> 2) What is the segment layout of the LVM PVs and LVs?
I am a bit lost with it. Is that what you are asking about?
master:
# pvs --segments
  PV         VG        Fmt  Attr PSize  PFree  Start SSize
  /dev/sda5  ubuntu-vg lvm2 a--  19.76g 20.00m     0  4926
  /dev/sda5  ubuntu-vg lvm2 a--  19.76g 20.00m  4926   127
  /dev/sda5  ubuntu-vg lvm2 a--  19.76g 20.00m  5053     5
# lvs --segments
  LV     VG        Attr      #Str Type   SSize  
  root   ubuntu-vg -wi-ao---    1 linear  19.24g
  swap_1 ubuntu-vg -wi-ao---    1 linear 508.00m


slave:
# pvs --segments
  PV         VG                  Fmt  Attr PSize   PFree Start SSize
  /dev/sda3  postgresnlpslave-vg lvm2 a--  429.77g    0      0 110021
  /dev/sda5  postgresnlpslave-vg lvm2 a--  169.52g    0      0  28392
  /dev/sda5  postgresnlpslave-vg lvm2 a--  169.52g    0  28392   2199
  /dev/sda5  postgresnlpslave-vg lvm2 a--  169.52g    0  30591   2560
  /dev/sda5  postgresnlpslave-vg lvm2 a--  169.52g    0  33151  10246
  /dev/sdb1  postgresnlpslave-vg lvm2 a--  512.00g    0      0 131071
# lvs --segments
  LV     VG                  Attr       #Str Type   SSize  
  root   postgresnlpslave-vg -wi-ao----    1 linear 110.91g
  root   postgresnlpslave-vg -wi-ao----    1 linear  40.02g
  root   postgresnlpslave-vg -wi-ao----    1 linear  10.00g
  root   postgresnlpslave-vg -wi-ao----    1 linear 429.77g
  root   postgresnlpslave-vg -wi-ao----    1 linear 512.00g
  swap_1 postgresnlpslave-vg -wi-ao----    1 linear   8.59g


> 3) Do you use LVM for any "extra" features, such as snapshots?
I don't think so, but how to check? vgs gives #SN = 0, is that it?

> 4) You can try using seekwatcher to see where on the disk the slowness is occurring. You get a chart similar to this http://kernel.dk/dd-md0-xfs-pdflush.png
> 5) BCC is a collection of tools that might shed a light on what is happening. https://github.com/iovisor/bcc
Will look into it.

Regards,
Vlad

pgsql-performance by date:

Previous
From: Alvaro Herrera
Date:
Subject: Re: link to Slow_Query_Questions from wiki/Main Page
Next
From: David Rowley
Date:
Subject: Re: To keep indexes in memory, is large enough effective_cache_size enough?