1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
|
What We Run, and Why
====================
Expressed as compactly as possible, the Provisioner is responsible for making
sure that non-Dead machine entities in state have agents running on live
instances; and for making sure that Dead machines, and stray instances, are
removed and cleaned up.
However, the choice of exactly what we deploy involves some subtleties. At the
Provisioner level, it's simple: the series and the constraints we pass to the
Environ.StartInstance come from the machine entity. But how did they get there?
Series
------
Individual charms are released for different possible target series; juju
should guarantee that charms for series X are only ever run on series X.
Every service, unit, and machine has a series that's set at creation time and
subsequently immutable. Units take their series from their service, and can
only be assigned to machines with matching series.
Subordinate units cannot be assigned directly to machines; they are created
by their principals, on the same machine, in response to the creation of
subordinate relations. We therefore restrict subordinate relations such that
they can only be created between services with matching series.
Constraints
-----------
Constraints are stored for environments, services, units, and machines, but
unit constraints are not currently exposed because they're not needed outside
state, and are likely to just cause trouble and confusion if we expose them.
From the point of a user, there are environment constraints and service
constraints, and sensible manipulations of them lead to predictable unit
deployment decisions. The mechanism is as follows:
* when a unit is added, the current environment and service constraints
are collapsed into a single value and stored for the unit. (To be clear:
at the moment the unit is created, the current service and environment
constraints will be combined such that every constraint not set on the
service is taken from the environment (or left unset, if not specified
at all).
* when a machine is being added in order to host a given unit, it copies
its constraints directly from the unit.
* when a machine is being added without a unit associated -- for example,
when adding additional state servers -- it copies its constraints directly
from the environment.
In this way the following sequence of operations becomes predictable:
$ juju deploy --constraints mem=2G wordpress
$ juju set-constraints --service wordpress mem=3G
$ juju add-unit wordpress -n 2
...in that exactly one machine will be provisioned with the first set of
constraints, and exactly two of them will be provisioned using the second
set. This is much friendlier to the users than delaying the unit constraint
capture and potentially suffering subtle and annoying races.
Subordinate units cannot have constraints, because their deployment is
controlled by their principal units. There's only ever one machine to which
that subordinate could (and must) be deployed, and to restrict that further
by means of constraints will only confuse people.
Machine Status and Provisioning Errors (current)
------------------------------------------------
In the light of time pressure, a unit assigned to a machine that has not been
provisioned can be removed directly by calling `juju destroy-unit`. Any
provisioning error can thus be "resolved" in an unsophisticated but moderately
effective way:
$ juju destroy-unit borken/0
...in that at least broken units don't clutter up the service and prevent its
removal. However:
$ juju destroy-machine 1
...does not yet cause an unprovisioned machine to be removed from state (whether
directly, or indirectly via the provisioner; the best place to implement this
functionality is not clear).
Machine Status and Provisioning Errors (WIP)
--------------------------------------------
[TODO: figure this out; not yet implemented, somewhat speculative... in
particular, use of "resolved" may be inappropriate. Consider adding a
"retry" CLI tool...]
When the provisioner fails to start a machine, it should ensure that (1) the
machine has no instance id set and (2) the machine has an error status set
that communicates the nature of the problem. This must be visible in the
output of `juju status`; and we must supply suitable tools to the user so
as to allow her to respond appropriately.
If the user believes a machine's provisioning error to be transient, she can
do a simple `juju resolved 14` which will set some state to make machine 14
eligible for the provisioner's attention again.
It may otherwise be that the unit ended up snapshotting a service/environ
config pair that really isn't satsifiable. In that case, the user can try
(say) `juju resolved 14 --constraints "mem=2G cpu-power=400"`, which allows
her to completely replace the machine's constraints as well as marking the
machine for reprovisioning attention.
|