Statistics
| Branch: | Tag: | Revision:

root / man / hsqueeze.rst @ e715a6d6

History | View | Annotate | Download (3 kB)

1
HSQUEEZE(1) Ganeti | Version @GANETI_VERSION@
2
=============================================
3

    
4
NAME
5
----
6

    
7
hsqueeze \- Dynamic power management
8

    
9
SYNOPSIS
10
--------
11

    
12
**hsqueeze** {backend options...} [algorithm options...] [reporting options...]
13

    
14
**hsqueeze** \--version
15

    
16
Backend options:
17

    
18
{ **-L[** *path* **]** | **-t** *data-file* }
19

    
20
Algorithm options:
21

    
22
**[ \--minimal-resources=*factor* ]**
23
**[ \--target-resources=*factor* ]**
24

    
25
Reporting options:
26

    
27
**[ -S *file* ]**
28
**[ -C[*file*] ]**
29

    
30

    
31
DESCRIPTION
32
-----------
33

    
34
hsqueeze does dynamic power management, by powering up or shutting down nodes,
35
depending on the current load of the cluster. Currently, only suggesting nodes
36
is implemented.
37

    
38
ALGORITHM
39
~~~~~~~~~
40

    
41
hsqueeze considers all online non-master nodes with only externally mirrored
42
instances as candidates for being taken offline. These nodes are iteratively,
43
starting from the node with the least number of instances, added to the set
44
of nodes to be put offline, if possible. A set of nodes is considered as suitable
45
for being taken offline, if, after marking these nodes as offline, balancing the
46
cluster by the algorithm used by **hbal**\(1) yields a situation where all instances
47
are located on online nodes, and each node has at least the target resources free
48
for new instances.
49

    
50
All offline nodes with a tag starting with ``htools:standby`` are
51
considered candidates for being taken online. Those nodes are taken online
52
till balancing the cluster by the algorithm used by **hbal**\(1) yields a
53
situation where each node has at least the minimal resources free for new
54
instances.
55

    
56
OPTIONS
57
-------
58

    
59
-L [*path*]
60
  Backend specification: collect data directly from the master daemon,
61
  which is to be contacted via LUXI (an internal Ganeti protocol). The
62
  option is described in the man page **htools**\(1).
63

    
64
-S *filename*, \--save-cluster=*filename*
65
  If given, the state of the cluster before the squeezing is saved to
66
  the given file plus the extension "original"
67
  (i.e. *filename*.original), and the state at the end of the
68
  squeezing operation is saved to the given file plus the extension "squeezed"
69
  (i.e. *filename*.squeezed).
70

    
71
-C[*filename*], \--print-commands[=*filename*]
72
  If given, a shell script containing the commands to squeeze or unsqueeze
73
  the cluster are saved in a file with the given name; if no name is provided,
74
  they are printed to stdout.
75

    
76
-t *datafile*, \--text-data=*datafile*
77
  Backend specification: the name of the file holding node and instance
78
  information (if not collecting LUXI). This or one of the
79
  other backends must be selected. The option is described in the man
80
  page **htools**\(1).
81

    
82
\--minimal-resources=*factor*
83
  Specify the amount of resources to be free on each node for hsqueeze not to
84
  consider onlining additional nodes. The value is reported a multiple of the
85
  standard instance specification, as taken from the instance policy.
86

    
87
\--target-resources=*factor*
88
  Specify the amount of resources to remain free on any node after squeezing.
89
  The value is reported a multiple of the standard instance specification, as
90
  taken from the instance policy.