mjf

Reclaiming space from bloated table without enough space left

The vacuum full needs two-times the table size to perform on it because it first makes full physical copy of the table. If you are running out of space and in need of reclaiming the space occupated by dead tuples from a bloated database table but you can not because of lacking free space in the table's namespace here is how you can work around that (assuming you some additional disk space resources can be added to the server). It is also the least disruptive method of doing it (in comparison to the obvious dump and restore et cetera).

mjf

Year	List	Subject:	Version
2017	pgsql-hackers	VACUUM and ANALYZE disagreeing on what reltuples means	<= 10

mjf

Personal

• Niko Matsakis - excellent and must read articles (!)

mjf

3x run?! One rights after another?!

zimbra@zimbra:~$ zmprov modifyConfig zimbraMtaMaxMessageSize 1073741824
/opt/zimbra/bin/zmjava: line 59: /bin/java: No such file or directory
zimbra@zimbra:~$ zmprov modifyConfig zimbraMtaMaxMessageSize 1073741824
Killed
zimbra@zimbra:~$ zmprov modifyConfig zimbraMtaMaxMessageSize 1073741824

mjf

set background=dark

hi clear
if exists('syntax_on')
 syntax reset
endif

let g:colors_name = 'phoenix'

hi Normal ctermfg=250

mjf

                       Example Calculation
+------------------------------------------+--------------------+
|                          key_buffer_size |          16.000 MB |
|                         query_cache_size |          16.000 MB |
|                  innodb_buffer_pool_size |          64.000 MB |
|          innodb_additional_mem_pool_size |           8.000 MB |
|                   innodb_log_buffer_size |           8.000 MB |
+------------------------------------------+--------------------+
|                              BASE MEMORY |         112.000 MB |

mjf

# TODO: store just basename of the logfiles (in Starlark may be easy)
# TODO: include more information in tags (at least user_name and application_name?)

#
# INPUTS
#
[[inputs.tail]]

    name_override = "postgres_sql_state_code"

mjf

Postgres random articles

Syntax

• SQL Style Guide

PM and DR

• Transaction ID Wraparound in Postgres

Scaling

mjf

# Telegraf configuration to collect per-CPU cycles and instructions
# using perf(1).  You can then calculate CPI / IPC from the metrics.

# If Telegraf is not running as UID 0 (aka "root") you can either use
# sudo(1) with the perf(1) command, raise Telegraf capability, or set
# kernel.perf_event_paranoid = 0.  But be warned: "Hic sunt dracones".

[[inputs.exec]]

  name_prefix = "cpu_"

mjf

# Telegraf configuration to collect Pressure Stall Information (PSI)

[[inputs.file]]
  name_override = "pressure_cpu"
  # interval = "5s"
  files = ["/proc/pressure/cpu"]
  data_format = "grok"
  grok_patterns = [
    "%{NOTSPACE:type:tag} avg10=%{NUMBER:avg10:float} avg60=%{NUMBER:avg60:float} avg300=%{NUMBER:avg300:float} total=%{NUMBER:total:int}"
  ]