Jobs¶
Jobs are designed to represent asynchronous tasks that your cozy can execute. These tasks can be scheduled in advance, recurring or sudden and provide various services.
At the time, we do not provide “generic” and programmable tasks. This list of available workers is part of the defined API.
The job queue can be either local to the cozy-stack when used as a monolithic instance (self-hosted for instance) or distributed via a redis server for distributed infrastructures.
This doc introduces two cozy types:
io.cozy.jobs
for jobsio.cozy.triggers
for triggers
Triggers¶
Jobs can be queued up via the /jobs/queue/:worker-type
API, for a direct
execution. But it can also be convenient to schedule jobs on some conditions,
and the triggers are the way to do that.
Jobs can be launched by different types of triggers:
@at
to schedule a one-time job executed after at a specific time in the future@in
to schedule a one-time job executed after a specific amount of time@hourly
to schedule jobs that run once an hour@daily
to schedule jobs that run once a day@weekly
to schedule jobs that run once a week@monthly
to schedule jobs that run once a month@every
to schedule periodic jobs executed at a given fix interval@cron
to schedule recurring jobs scheduled at specific times@event
to launch a job after a change on documents in the cozy@webhook
to launch a job when an HTTP request hit a specific URL@client
when the client controls when the job are launched.
These triggers have specific syntaxes to describe when jobs should be scheduled. See below for more informations.
@at
syntax¶
The @at
trigger takes a ISO-8601 formatted string indicating a UTC time in the
future. The date is of this form: YYYY-MM-DDTHH:mm:ss.sssZ
Be aware that the @at
trigger is removed from the doctype after it has created the associated job.
Examples
@at 2018-12-12T15:36:25.507Z
@in
syntax¶
The @in
trigger takes the same duration syntax as @every
Be aware that the @in
trigger is removed from the doctype after it has created the associated job.
Examples
@in 10m @in 1h30m
@daily
syntax¶
The @daily
trigger will create a job once a day. By default, the stack is
free to choose the day and hour when to do that, but you can add restrictions:
@daily # Once a day, any hour @daily before 5am # Once a day, between midnight and 5am (UTC) @daily after 10pm # Once a day, between 10pm and midnight (UTC) @daily between 8am and 6pm # Once a day, between 8am and 6pm (UTC)
@weekly
syntax¶
The @weekly
trigger will create a job once a week. By default, the stack is
free to choose the day and hour when to do that, but you can add restrictions:
@weekly # Once a week, any day, any hour @weekly on monday # Every monday @weekly on mon,wed,fri # Once a week, on a monday, wednesday, or friday @weekly on wed-fri # Once a week, on a wednesday, thursday, or friday @weekly on weekday # Once a week, but not the week-end @weekly on weekend # Every week-end (saturday or sunday) @weekly before 5am # Once a week, any day, but between midnight and 5am (UTC) @weekly after 10pm # Once a week, any day, but between 10pm and midnight (UTC) @weekly between 8am and 6pm # Once a week, any day, between 8am and 6pm (UTC) @weekly on monday before 9am # Every monday, between midnight and 9am (UTC)
@monthly
syntax¶
The @monthly
trigger will create a job once a month. By default, the stack is
free to choose the day and hour when to do that, but you can add restrictions:
@monthly # Once a month, any day, any hour @monthly on the 1 # Every first day of the month @monthly on the 1-5 # Once a month, on the first five days of the month @monthly before 5am # Once a month, any day, but between midnight and 5am (UTC) @monthly after 10pm # Once a month, any day, but between 10pm and midnight (UTC) @monthly between 8am and 6pm # Once a month, any day, between 8am and 6pm (UTC) @monthly on the 1 before 9am # Every first day of the month, between midnight and 9am (UTC)
Note: the current implementation is to take a random day/hour and run the job each month at this day/hour. So, you should avoid 29-31 if you really want the job to run each month.
@every
syntax¶
The @every
trigger uses the same syntax as golang’s time.ParseDuration
(but
only support time units above seconds):
A duration string is a possibly signed sequence of decimal numbers, each with optional fraction and a unit suffix, such as “300ms”, “1.5h” or “2h45m”. Valid time units are “s”, “m”, “h”.
Examples
@every 1.5h # schedules every 1 and a half hours @every 30m10s # schedules every 30 minutes and 10 seconds
@cron
syntax¶
In order to schedule recurring jobs, the @cron
trigger has the syntax using
six fields:
Field name | Mandatory? | Allowed values | Allowed special characters |
---|---|---|---|
Seconds | Yes | 0-59 | * / , - |
Minutes | Yes | 0-59 | * / , - |
Hours | Yes | 0-23 | * / , - |
Day of month | Yes | 1-31 | * / , - ? |
Month | Yes | 1-12 or JAN-DEC | * / , - |
Day of week | Yes | 0-6 or SUN-SAT | * / , - ? |
Asterisk ( *
)
The asterisk indicates that the cron expression will match for all values of the field; e.g., using an asterisk in the 5th field (month) would indicate every month.
Slash ( /
)
Slashes are used to describe increments of ranges. For example 3-59/15 in the
1st field (minutes) would indicate the 3rd minute of the hour and every 15
minutes thereafter. The form "*\/..."
is equivalent to the form
"first-last/..."
, that is, an increment over the largest possible range of the
field. The form "N-/..."
is accepted as meaning "N-MAX/..."
, that is,
starting at N, use the increment until the end of that specific range. It does
not wrap around.
Comma ( ,
)
Commas are used to separate items of a list. For example, using "MON,WED,FRI"
in the 5th field (day of week) would mean Mondays, Wednesdays and Fridays.
Hyphen ( -
)
Hyphens are used to define ranges. For example, 9-17 would indicate every hour between 9am and 5pm inclusive.
Question mark ( ?
)
Question mark may be used instead of *
for leaving either day-of-month or
day-of-week blank.
To schedule jobs given an interval:
Examples:
@cron 0 0 0 1 1 * # Run once a year, midnight, Jan. 1st @cron 0 0 0 1 1 * # Run once a year, midnight, Jan. 1st @cron 0 0 0 1 * * # Run once a month, midnight, first of month @cron 0 0 0 * * 0 # Run once a week, midnight on Sunday @cron 0 0 0 * * * # Run once a day, midnight @cron 0 0 * * * * # Run once an hour, beginning of hour
@event
syntax¶
The @event
syntax allows to trigger a job when something occurs in the stack.
It follows the same syntax than permissions
scope
string:
type[:verb][:values][:selector]
Unlike for permissions string, the verb should be one of CREATED
, DELETED
,
UPDATED
. It is possible to put several verbs, separated by a comma.
There is also a special value !=
. It means that a job will be trigger only if
the value for the given selector has changed (ie the value before the update and
the value after that are different).
The job worker will receive a compound message including original trigger_infos messages and the event which has triggered it.
Examples:
@event io.cozy.files // anything happens on files @event io.cozy.files:CREATED // a file was created @event io.cozy.files:DELETED:image/jpg:mime // an image was deleted @event io.cozy.bank.operations:CREATED io.cozy.bank.bills:CREATED // a bank operation or a bill @event io.cozy.bank.operations:CREATED,UPDATED // a bank operation created or updated @event io.cozy.bank.operations:UPDATED:!=:category // a change of category for a bank operation
@webhook
syntax¶
It takes no parameter. The URL to hit is not controlled by the request, but is
chosen by the server (and is returned as webhook
in the links
JSON-API
response).
Example:
@webhook
@client
syntax¶
It takes no parameter and can only by used for the client
worker. The stack
won’t create a job unless a client calls the launch endpoint for this trigger.
The main goal of this trigger is keep a state, as the aggregation of job
results.
Error Handling¶
Jobs can fail to execute their task. We have two ways to parameterize such cases.
Retry¶
A retry count can be optionally specified to ask the worker to re-execute the task if it has failed.
Each retry is executed after a configurable delay. The try count is part of the
attributes of the job. Also, each occurring error is kept in the errors
field
containing all the errors that may have happened.
Timeout¶
A worker may never end. To prevent this, a configurable timeout value is specified with the job.
If a job does not end after the specified amount of time, it will be aborted. A timeout is just like another error from the worker and can provoke a retry if specified.
Defaults¶
By default, jobs are parameterized with a maximum of 3 tries with 1 minute timeout.
These defaults may vary given the workload of the workers.
Jobs API¶
Example and description of the attributes of a io.cozy.jobs
:
{ "domain": "me.cozy.localhost", "worker": "sendmail", // worker type name "options": { "timeout": 60, // timeout value in seconds "max_exec_count": 3, // maximum number of time the job should be executed (including retries) }, "arguments": { // the arguments will be given to the worker (if you look in CouchDB, it is called message there) "mode": "noreply", "template_name": "new_registration", "template_values": { "DevicesLink": "http://me.cozy.localhost/#/connectedDevices", } }, "state": "running", // queued, running, done, errored "queued_at": "2016-09-19T12:35:08Z", // time of the queuing "started_at": "2016-09-19T12:35:08Z", // time of first execution "error": "" // error message if any }
Example and description of a job creation options — as you can see, the options
are replicated in the io.cozy.jobs
attributes:
{ "timeout": 60, // timeout value in seconds "max_exec_count": 3, // maximum number of retry }
GET /jobs/:job-id¶
Get a job informations given its ID.
Request¶
GET /jobs/123123 HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": { "type": "io.cozy.jobs", "id": "123123", "attributes": { "domain": "me.cozy.localhost", "worker": "sendmail", "options": { "timeout": 60, "max_exec_count": 3 }, "state": "running", "queued_at": "2016-09-19T12:35:08Z", "started_at": "2016-09-19T12:35:08Z", "error": "" }, "links": { "self": "/jobs/123123" } } }
POST /jobs/queue/:worker-type¶
Enqueue programmatically a new job.
This route requires a specific permission on the worker-type. A global
permission on the global io.cozy.jobs
doctype is not allowed.
Each worker accepts different arguments. For konnectors, the
arguments will be given in the process.env['COZY_FIELDS']
variable.
Request¶
POST /jobs/queue/sendmail HTTP/1.1 Content-Type: application/vnd.api+json Accept: application/vnd.api+json
{ "data": { "attributes": { "manual": false, "options": { "timeout": 60, "max_exec_count": 3 }, "arguments": {} // any json value used as arguments for the job } } }
Response¶
{ "data": { "type": "io.cozy.jobs", "id": "123123", "attributes": { "domain": "me.cozy.localhost", "worker": "sendmail", "options": { "timeout": 60, "max_exec_count": 3 }, "state": "running", "queued_at": "2016-09-19T12:35:08Z", "started_at": "2016-09-19T12:35:08Z", "error": "" }, "links": { "self": "/jobs/123123" } } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.jobs
for the verb POST
. The is required to restrict its permission
to specific worker(s), like this (a global permission on the doctype is not
allowed):
{ "permissions": { "mail-from-the-user": { "description": "Required to send mails from the user to his/her friends", "type": "io.cozy.jobs", "verbs": ["POST"], "selector": "worker", "values": ["sendmail"] } } }
POST /jobs/support¶
Send a mail to the support (email address defined by mail.reply_to
in the
config file, or overwritten by context with contexts.<name>.reply_to
).
It requires a permission on io.cozy.support
(a permission on
io.cozy.jobs:POST:sendmail:worker
, or larger, is also accepted to ease the
transition from sending manually a mail to the support via the sendmail queue).
Request¶
POST /jobs/support HTTP/1.1 Content-Type: application/vnd.api+json
{ "data": { "attributes": { "arguments": { "subject": "Cozy is so cool!", "body": "I really love Cozy. Thank you so much!" } } } }
Response¶
HTTP/1.1 204 No Content
POST /jobs/campaign-emails¶
Send a non transactional (or campaign) email to the user via the dedicated
campaign mail server (configured via campaign_mail
attributes in the config
file, or overwritten by context with campaign_mail.contexts.<name>
attributes).
Both the subject and at least one part are required.
Request¶
POST /jobs/campaign-emails HTTP/1.1 Content-Type: application/vnd.api+json
{ "data": { "attributes": { "arguments": { "subject": "Checkout the new cool stuff!", "parts": [ { "body": "So many new features to check out!", "type": "text/plain" } ] } } } }
Response¶
HTTP/1.1 204 No Content
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.jobs
for the verb POST
and the sendmail
worker.
This can be defined like so:
{ "permissions": { "campaign-emails": { "description": "Required to send campaign emails to the user", "type": "io.cozy.jobs", "verbs": ["POST"], "selector": "worker", "values": ["sendmail"] } } }
GET /jobs/queue/:worker-type¶
List the jobs in the queue.
Request¶
GET /jobs/queue/sendmail HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": [ { "attributes": { "domain": "cozy.localhost:8080", "options": null, "queued_at": "2017-09-29T15:32:31.953878568+02:00", "started_at": "0001-01-01T00:00:00Z", "state": "queued", "worker": "log" }, "id": "77689bca9634b4fb08d6ca3d1643de5f", "links": { "self": "/jobs/log/77689bca9634b4fb08d6ca3d1643de5f" }, "meta": { "rev": "1-f823bcd2759103a5ad1a98f4bf083b36" }, "type": "io.cozy.jobs" } ], "meta": { "count": 0 } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.jobs
for the verb GET
. In most cases, the application will restrict
its permission to only one worker, like this:
{ "permissions": { "mail-from-the-user": { "description": "Required to know the number of jobs in the sendmail queues", "type": "io.cozy.jobs", "verbs": ["GET"], "selector": "worker", "values": ["sendmail"] } } }
PATCH /jobs/:job-id¶
This endpoint can be used for a job of the client
worker (executed by a
client, not on the server) to update the status.
Request¶
PATCH /jobs/022368c07dc701396403543d7eb8149c HTTP/1.1 Content-Type: application/vnd.api+json
{ "data": { "type": "io.cozy.jobs", "id": "022368c07dc701396403543d7eb8149c", "attributes": { "state": "errored", "error": "LOGIN_FAILED" } } }
Response¶
HTTP/1.1 200 OK Content-Type: application/vnd.api+json
{ "data": { "type": "io.cozy.jobs", "id": "022368c07dc701396403543d7eb8149c", "attributes": { "domain": "me.cozy.localhost", "worker": "sendmail", "options": {}, "state": "errored", "error": "LOGIN_FAILED", "queued_at": "2021-04-12T12:34:56Z", "started_at": "2021-04-12T12:34:56Z", "finished_at": "2021-04-12T12:38:59Z" }, "links": { "self": "/jobs/022368c07dc701396403543d7eb8149c" } } }
POST /jobs/triggers¶
Add a trigger of the worker. See triggers’ descriptions to see the types of trigger and their arguments syntax.
This route requires a specific permission on the worker type. A global
permission on the global io.cozy.triggers
doctype is not allowed.
The debounce
parameter can be used to limit the number of jobs created in a
burst. It delays the creation of the job on the first input by the given time
argument, and if the trigger has its condition matched again during this period,
it won’t create another job. It can be useful to combine it with the changes
feed of couchdb with a last sequence number persisted by the worker, as it
allows to have a nice diff between two executions of the worker. Its syntax is the
one understood by go’s time.ParseDuration.
Request¶
POST /jobs/triggers HTTP/1.1 Accept: application/vnd.api+json
{ "data": { "attributes": { "type": "@event", "arguments": "io.cozy.invitations", "debounce": "10m", "worker": "sendmail", "message": {}, "options": { "timeout": 60, "max_exec_count": 3 } } } }
Note: the message
field was previously called worker_arguments
. The
latter version still works but is deprecated, you should use message
instead.
Response¶
{ "data": { "type": "io.cozy.triggers", "id": "123123", "attributes": { "type": "@every", "arguments": "30m10s", "debounce": "10m", "worker": "sendmail", "options": { "timeout": 60, "max_exec_count": 3 } }, "links": { "self": "/jobs/triggers/123123" } } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb POST
. In most cases, the application will
restrict its permission to only one worker, like this:
{ "permissions": { "mail-from-the-user": { "description": "Required to send regularly mails from the user to his/her friends", "type": "io.cozy.triggers", "verbs": ["POST"], "selector": "worker", "values": ["sendmail"] } } }
GET /jobs/triggers/:trigger-id¶
Get a trigger informations given its ID.
Request¶
GET /jobs/triggers/123123 HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": { "type": "io.cozy.triggers", "id": "123123", "attributes": { "type": "@every", "arguments": "30m10s", "worker": "sendmail", "options": { "timeout": 60, "max_exec_count": 3 }, "current_state": { "status": "done", "last_success": "2017-11-20T13:31:09.01641731", "last_successful_job_id": "abcde", "last_execution": "2017-11-20T13:31:09.01641731", "last_executed_job_id": "abcde", "last_failure": "2017-11-20T13:31:09.01641731", "last_failed_job_id": "abcde", "last_error": "error value", "last_manual_execution": "2017-11-20T13:31:09.01641731", "last_manual_job_id": "abcde" } }, "links": { "self": "/jobs/triggers/123123" } } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb GET
. A konnector can also call this endpoint
for one of its triggers (no permission required).
GET /jobs/triggers/:trigger-id/state¶
Get the trigger current state, to give a big picture of the health of the trigger.
- last executed job status (
done
,errored
,queued
orrunning
) - last executed job that resulted in a successful executoin
- last executed job that resulted in an error
- last executed job from a manual execution (not executed by the trigger directly)
Request¶
GET /jobs/triggers/123123/state HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": { "type": "io.cozy.triggers.state", "id": "123123", "attributes": { "status": "done", "last_success": "2017-11-20T13:31:09.01641731", "last_successful_job_id": "abcde", "last_execution": "2017-11-20T13:31:09.01641731", "last_executed_job_id": "abcde", "last_failure": "2017-11-20T13:31:09.01641731", "last_failed_job_id": "abcde", "last_error": "error value", "last_manual_execution": "2017-11-20T13:31:09.01641731", "last_manual_job_id": "abcde" } } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb GET
. A konnector can also call this endpoint
for one of its triggers (no permission required).
GET /jobs/triggers/:trigger-id/jobs¶
Get the jobs launched by the trigger with the specified ID.
Query parameters:
Limit
: to specify the number of jobs to get out
Request¶
GET /jobs/triggers/123123/jobs?Limit=1 HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": [ { "type": "io.cozy.jobs", "id": "123123", "attributes": {}, "links": { "self": "/jobs/123123" } } ] }
PATCH /jobs/triggers/:trigger-id¶
This route can be used to change the frequency of execution of a @cron
trigger, or the message of any trigger.
Request¶
PATCH /jobs/triggers/123123 HTTP/1.1 Content-Type: application/vnd.api+json Accept: application/vnd.api+json
{ "data": { "attributes": { "type": "@cron", "arguments": "0 0 0 * * 0" } } }
Response¶
{ "data": { "type": "io.cozy.triggers", "id": "123123", "attributes": { "type": "@cron", "arguments": "0 0 0 * * 0", "worker": "sendmail", "options": { "timeout": 60, "max_exec_count": 3 } }, "links": { "self": "/jobs/triggers/123123" } } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb PATCH
. A konnector can also call this
endpoint for one of its triggers (no permission required).
POST /jobs/triggers/:trigger-id/launch¶
Launch a trigger manually given its ID and return the created job.
Note: this endpoint can be used to create a job for a @client
trigger. In
that case, the job won’t be executed on the server but by the client. And the client
must call PATCH /jobs/:job-id
when the job is completed (success or error).
Request¶
POST /jobs/triggers/123123/launch HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": { "type": "io.cozy.jobs", "id": "123123", "attributes": { "domain": "me.cozy.localhost", "worker": "sendmail", "options": {}, "state": "running", "queued_at": "2016-09-19T12:35:08Z", "started_at": "2016-09-19T12:35:08Z", "error": "" }, "links": { "self": "/jobs/123123" } } }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb POST
.
DELETE /jobs/triggers/:trigger-id¶
Delete a trigger given its ID.
Request¶
DELETE /jobs/triggers/123123 HTTP/1.1 Accept: application/vnd.api+json
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb DELETE
. A konnector can also call this
endpoint for one of its triggers (no permission required).
GET /jobs/triggers¶
Get the list of triggers. This route only accept Worker
and Type
query
parameters and returns the trigger but also in attributes
its current_state
(the same current_state
returned by GET
/jobs/triggers/:trigger-id). Be warned that
/data/io.cozy.triggers/_find
does not return this current_state
attribute
and you’ll need to query /jobs/triggers/:trigger-id
to have it.
Query parameters (with comma-separated values):
Type
: to filter on the trigger type (@cron
,@in
, etc.)Worker
: to filter only triggers associated with a specific worker.
Request¶
GET /jobs/triggers?Worker=konnector&Type=@cron,@in,@at HTTP/1.1 Accept: application/vnd.api+json
Response¶
{ "data": [ { "type": "io.cozy.triggers", "id": "123123", "arguments": "0 40 0 * * *", "current_state": { "last_error": "LOGIN_FAILED", "last_executed_job_id": "abc", "last_execution": "2019-01-07T08:23:22.069902888Z", "last_failed_job_id": "abcd", "last_failure": "2019-01-07T08:23:22.069902888Z", "last_manual_execution": "2019-01-07T08:23:22.069902888Z", "last_manual_job_id": "azer", "status": "errored", "trigger_id": "123123" }, "debounce": "", "domain": "xxx.mycozy.cloud", "message": { "konnector": "slug", "account": "XXX" }, "options": null, "type": "@cron", "worker": "konnector", "id": "123123", "links": { "self": "/jobs/triggers/123123" } } ] }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.triggers
for the verb GET
. When used on a specific worker, the
permission can be specified on the worker
field.
POST /jobs/webhooks/:trigger-id¶
This endpoint is used for creating a job (for example executing a konnector
or a service).
It requires no permission, but a trigger of type @webhook
must have been
created before using this endpoint. Its body must be a JSON that will be
available to the konnector or to the service through the
process.env['COZY_PAYLOAD']
variable.
It is possible to pass a Manual=true
parameter in the query-string if the job
is interactive. It will give it an higher priority in the queues.
Request¶
POST /jobs/webhooks/f34c74d0-0c91-0139-5af5-543d7eb8149c HTTP/1.1 Content-Type: application/json
{ "account_id": "0672e560" }
Response¶
HTTP/1.1 204 No Content
POST /jobs/webhooks/bi¶
This endpoint is used to create jobs for banking konnectors. It requires a
payload with the format defined by Budget
Insight and an
Authorization
header with a Bearer token, where a trigger and an account can
be found on this instance matching their data. The event type and the URL of
the BI API (on the good environment) are also sent in the query string.
Request¶
POST /jobs/webhooks/bi?event=CONNECTION_SYNCED&bi_url=https://.../ HTTP/1.1 Content-Type: application/json Authorization: Bearer token-from-bi Host: cozy.example.com
Response¶
HTTP/1.1 204 No Content
DELETE /jobs/purge¶
This endpoint allows to purge old jobs of an instance. Some parameters can be given to this route:
duration
is the duration of jobs to keep. This is a human-readable string (integer+suffix). For example:- “3W” will keep jobs up to 3 weeks
- “2M” will keep jobs up to 2 months
workers
is a comma-separated list of workers to apply the purge job.
Request¶
DELETE /jobs/purge HTTP/1.1 Accept: application/json
Response¶
{ "deleted": 42 }
Permissions¶
To use this endpoint, an application needs a permission on the type
io.cozy.jobs
for the verb DELETE
.
Worker pool¶
The consuming side of the job queue is handled by a worker pool.
On a monolithic cozy-stack, the worker pool has a configurable fixed size of workers. The default value is not yet determined. Each time a worker has finished a job, it check the queue and based on the priority and the queued date of the job, picks a new job to execute.
Permissions¶
In order to prevent jobs from leaking informations between applications, we may need to add filtering per applications: for instance one queue per applications.
We should also have an explicit check on the permissions of the applications before launching a job scheduled by an application. For more information, refer to our permission document.
Multi-stack¶
When some instances are served by several stacks, the scheduling and running of jobs can be distributed on the stacks. The synchronization is done via redis.
For scheduling, there is one important key in redis: triggers
. It’s a sorted
set. The members are the identifiants of the triggers (with the domain name),
and the score are the timestamp of the next time the trigger should occur.
During the short period of time where a trigger is processed, its key is moved
to scheduling
(another sorted set). So, even if a stack crash during
processing a trigger, this trigger won’t be lost.
For @event
triggers, we don’t use the same mechanism. Each stack has all the
triggers in memory and is responsible to trigger them for the events generated
by the HTTP requests of their API. They also publish them on redis: this pub/sub
is used for the realtime API.