Newer
Older
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
107
# Task service - Task list
### Task list definition
A task list is described in JSON format and base task list (template) definition *must* be available
before a task list can be created. Task list JSON templates are stored in Mongo database in a collection
named `taskListTemplates`. Below is an example of task list template definition:
```json
{
"name": "example",
"cacheNamespace": "login",
"cacheScope": "user",
"groups": [
{
"execution": "sequential",
"tasks": [
"taskName1",
"taskName2"
]
},
{
"execution": "parallel",
"tasks": [
"taskName3",
"taskName4",
"taskName5"
]
}
]
}
```
Task lists are created by using their `name` attribute. Tasks for each group in the `groups` field
are also created by using their `name` so task names should match existing task template
definitions (see [tasks](task.md)). If a task list template with the given `name` does not exist OR a task
template definition does not exist, a task list will not be created and an error is returned.
### Task list Execution
Below is an example of creating a task list with the template definition given above:
```shell
curl -v -X POST http://localhost:8082/v1/taskList/example -d '{"input": {"key": "value"}}'
```
The HTTP request will create a task list and corresponding tasks for each of the groups within the task list
for asynchronous execution and the JSON object given as input will be used as the body of the task list request.
The caller will receive immediately the `taskListID` as response.
The executor then takes a task list from the queue and executes the groups within the task list sequentially.
Each group has a field called `execution` which can be one of two options: `parallel` or `sequential`. This
field describes how the tasks within the group *must* be executed.
- Sequential group execution: tasks within the group are executed sequentially and the result of each task is passed as
an input to the next. If one task fails to execute, all following tasks are marked as failed and the whole group fails.
- Parallel group execution: tasks within the group are executed in parallel and the results are dependant. If a task
fails to execute, this does not affect the other tasks but the group is marked with failed status.
The state of the task list asynchronous execution is available later on the `result` endpoint:
```shell
curl -v -X GET http://localhost:8082/v1/taskListResult/{taskListID}
```
### Task list Executor Configuration
There are two environment variables that control the level of concurrency
of the task list executor.
```shell
LIST_EXECUTOR_WORKERS="5"
LIST_EXECUTOR_POLL_INTERVAL="1s"
```
Poll interval specifies how often the executor will try to fetch a *pending* task list
from the queue. After a task list is fetched, it is executed in parallel with other executions.
Number of workers is the limit of permitted parallel task lists executions.
With the given example of 1 second (1s), the executor will retrieve 1 task list per second at most.
If there are multiple instances (pods) of the service, multiply by their number
(e.g. 5 pods = 5 tasks per second).
If this is not enough, the poll interval can be decreased, or we can slightly modify
the polling function to fetch many task lists at once (and also increase the number of workers).
To learn more about the queue and why we use database as queue see [queue](queue.md)
### Task List Storage
We use MongoDB for task list storage. There are three Mongo collections with different purpose.
1. **taskListTemplates**
The collection contains predefined task list definitions in JSON format. Each definition contains
groups of tasks which must be instantiated and later executed as part of the task list.
2. **taskLists**
The collection contains newly created task lists *pending* for execution. It acts like a
FIFO queue and is used by the task list executor to retrieve task lists for workers to execute.
3. **tasks**
The collection contains the tasks belonging to a group which is part of a task list. When a task list
is fetched for execution, all tasks are fetched and executed for that particular task list.
4. **tasksListHistory**
The collection contains completed task lists for results querying,
audit, reporting and debugging purposes.