2018-10-22 03:00:50 -04:00
|
|
|
# frozen_string_literal: true
|
|
|
|
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
module Gitlab
|
|
|
|
# JobWaiter can be used to wait for a number of Sidekiq jobs to complete.
|
2017-08-15 08:56:04 -04:00
|
|
|
#
|
|
|
|
# Its use requires the cooperation of the sidekiq jobs themselves. Set up the
|
|
|
|
# waiter, then start the jobs, passing them its `key`. Their `perform` methods
|
|
|
|
# should look like:
|
|
|
|
#
|
|
|
|
# def perform(args, notify_key)
|
|
|
|
# # do work
|
|
|
|
# ensure
|
|
|
|
# ::Gitlab::JobWaiter.notify(notify_key, jid)
|
|
|
|
# end
|
|
|
|
#
|
|
|
|
# The JobWaiter blocks popping items from a Redis array. All the sidekiq jobs
|
|
|
|
# push to that array when done. Once the waiter has popped `count` items, it
|
|
|
|
# knows all the jobs are done.
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
class JobWaiter
|
2019-08-31 15:25:25 -04:00
|
|
|
KEY_PREFIX = "gitlab:job_waiter"
|
2018-02-26 06:17:50 -05:00
|
|
|
|
2020-02-25 16:09:23 -05:00
|
|
|
STARTED_METRIC = :gitlab_job_waiter_started_total
|
|
|
|
TIMEOUTS_METRIC = :gitlab_job_waiter_timeouts_total
|
|
|
|
|
2017-08-15 08:56:04 -04:00
|
|
|
def self.notify(key, jid)
|
|
|
|
Gitlab::Redis::SharedState.with { |redis| redis.lpush(key, jid) }
|
|
|
|
end
|
|
|
|
|
2018-02-26 06:17:50 -05:00
|
|
|
def self.key?(key)
|
|
|
|
key.is_a?(String) && key =~ /\A#{KEY_PREFIX}:\h{8}-\h{4}-\h{4}-\h{4}-\h{12}\z/
|
|
|
|
end
|
|
|
|
|
2020-02-25 16:09:23 -05:00
|
|
|
attr_reader :key, :finished, :worker_label
|
2017-10-13 12:50:36 -04:00
|
|
|
attr_accessor :jobs_remaining
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
|
2017-08-15 08:56:04 -04:00
|
|
|
# jobs_remaining - the number of jobs left to wait for
|
2017-10-13 12:50:36 -04:00
|
|
|
# key - The key of this waiter.
|
2020-02-25 16:09:23 -05:00
|
|
|
def initialize(jobs_remaining = 0, key = "#{KEY_PREFIX}:#{SecureRandom.uuid}", worker_label: nil)
|
2017-10-13 12:50:36 -04:00
|
|
|
@key = key
|
2017-08-15 08:56:04 -04:00
|
|
|
@jobs_remaining = jobs_remaining
|
|
|
|
@finished = []
|
2020-02-25 16:09:23 -05:00
|
|
|
@worker_label = worker_label
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
end
|
|
|
|
|
|
|
|
# Waits for all the jobs to be completed.
|
|
|
|
#
|
|
|
|
# timeout - The maximum amount of seconds to block the caller for. This
|
|
|
|
# ensures we don't indefinitely block a caller in case a job takes
|
|
|
|
# long to process, or is never processed.
|
2017-06-19 10:12:05 -04:00
|
|
|
def wait(timeout = 10)
|
2017-08-15 08:56:04 -04:00
|
|
|
deadline = Time.now.utc + timeout
|
2020-02-25 16:09:23 -05:00
|
|
|
increment_counter(STARTED_METRIC)
|
2017-08-15 08:56:04 -04:00
|
|
|
|
|
|
|
Gitlab::Redis::SharedState.with do |redis|
|
|
|
|
# Fallback key expiry: allow a long grace period to reduce the chance of
|
|
|
|
# a job pushing to an expired key and recreating it
|
|
|
|
redis.expire(key, [timeout * 2, 10.minutes.to_i].max)
|
|
|
|
|
|
|
|
while jobs_remaining > 0
|
|
|
|
# Redis will not take fractional seconds. Prefer waiting too long over
|
|
|
|
# not waiting long enough
|
|
|
|
seconds_left = (deadline - Time.now.utc).ceil
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
|
2017-08-15 08:56:04 -04:00
|
|
|
# Redis interprets 0 as "wait forever", so skip the final `blpop` call
|
|
|
|
break if seconds_left <= 0
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
|
2017-08-15 08:56:04 -04:00
|
|
|
list, jid = redis.blpop(key, timeout: seconds_left)
|
2020-02-25 16:09:23 -05:00
|
|
|
|
|
|
|
# timed out
|
|
|
|
unless list && jid
|
|
|
|
increment_counter(TIMEOUTS_METRIC)
|
|
|
|
break
|
|
|
|
end
|
2017-08-15 08:56:04 -04:00
|
|
|
|
|
|
|
@finished << jid
|
|
|
|
@jobs_remaining -= 1
|
|
|
|
end
|
|
|
|
|
|
|
|
# All jobs have finished, so expire the key immediately
|
|
|
|
redis.expire(key, 0) if jobs_remaining == 0
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
end
|
2017-08-15 08:56:04 -04:00
|
|
|
|
|
|
|
finished
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
end
|
2020-02-25 16:09:23 -05:00
|
|
|
|
|
|
|
private
|
|
|
|
|
|
|
|
def increment_counter(metric)
|
|
|
|
return unless worker_label
|
|
|
|
|
|
|
|
metrics[metric].increment(worker: worker_label)
|
|
|
|
end
|
|
|
|
|
|
|
|
def metrics
|
|
|
|
@metrics ||= {
|
|
|
|
STARTED_METRIC => Gitlab::Metrics.counter(STARTED_METRIC, 'JobWaiter attempts started'),
|
|
|
|
TIMEOUTS_METRIC => Gitlab::Metrics.counter(TIMEOUTS_METRIC, 'JobWaiter attempts timed out')
|
|
|
|
}
|
|
|
|
end
|
Fix race conditions for AuthorizedProjectsWorker
There were two cases that could be problematic:
1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
transaction it was possible for a job to run/complete before a
COMMIT; resulting in it either producing an error, or producing no
new data.
2. When scheduling jobs the code would not wait until completion. This
could lead to a user creating a project and then immediately trying
to push to it. Usually this will work fine, but given enough load it
might take a few seconds before a user has access.
The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).
This commit changes two things to deal with this:
1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
by scheduling using Rails' after_commit hook instead of doing so in
an arbitrary method.
2. When scheduling jobs the calling thread now waits for all jobs to
complete.
Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:
* sidekiq-status
* sidekiq_status
No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.
Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.
This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.
Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.
The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.
For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.
In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.
2017-01-22 12:22:02 -05:00
|
|
|
end
|
|
|
|
end
|