e3cbb2c82e
allocs fit checks if devices get oversubscribed
2018-11-07 10:33:22 -08:00
4f9b3ede87
Split device accounter and allocator
2018-11-07 10:32:03 -08:00
6fa893c801
affinities
2018-11-07 10:32:03 -08:00
feb83a2be3
assign devices
2018-11-07 10:32:03 -08:00
6d8bb3a7bd
Duplicate blocked evals cancelling improved
...
The old logic for cancelling duplicate blocked evaluations by job id had
the issue where the newer evaluation could have additional node classes
that it is (in)eligible for that we would not capture. This could make
it such that cluster state could change such that the job would make
progress but no evaluation was unblocked.
2018-11-07 10:08:23 -08:00
a6b714b81c
update preemption tests to use new node resource structs
...
also includes a fix to remove unnecessary subtraction of network mbits
2018-11-02 17:59:53 -05:00
b2b52b1ada
Merge pull request #4794 from hashicorp/f-preemption-systemjobs
...
Preemption for system jobs
2018-11-02 16:28:06 -05:00
56de32f363
Address more minor code review feedback
2018-11-02 16:26:34 -05:00
253a351532
Fix test setup
2018-11-02 16:06:25 -05:00
fba24e5a8a
dereference safely
2018-11-02 15:58:59 -05:00
d061678df7
Fix static port preemption to be device aware
2018-11-02 13:07:24 -05:00
4182444937
Handle static port preemption when there are multiple devices
...
Also added test case
2018-11-02 09:09:50 -05:00
fd60e66f86
Plumb alloc resource cache in a few more places.
...
also removed now unused method
2018-11-01 16:44:43 -05:00
78d635edca
More review comments
2018-11-01 16:36:11 -05:00
6e1023ba08
Cleaner way to exit early, and fixed a couple more places reading from alloc.Resources
2018-11-01 16:15:58 -05:00
b4dd26247f
review comments
2018-11-01 12:01:59 -05:00
d03201adf8
Fix formatting of allocation score metrics
2018-10-30 12:03:23 -05:00
f1c3eb2792
Introduce interface with multiple implementations for resource distance
2018-10-30 11:06:32 -05:00
047af5141e
refactor preemption code to use method recievers and setters for common fields
2018-10-30 11:06:32 -05:00
1a5421f5d7
more minor cleanup
2018-10-30 11:06:32 -05:00
0494a098ce
More style and readablity fixes from review
2018-10-30 11:06:32 -05:00
3910ba9bbd
Preempted allocations should be removed from proposed allocations
2018-10-30 11:06:32 -05:00
9dd76d83dc
comments
2018-10-30 11:06:32 -05:00
e6234e3cc5
fix end to end scheduler test to use new resource structs correctly
2018-10-30 11:06:32 -05:00
8807c25b11
Modify preemption code to use new style of resource structs
2018-10-30 11:06:32 -05:00
c1c1c230e4
Make preemption config a struct to allow for enabling based on scheduler type
2018-10-30 11:06:32 -05:00
25a047267f
Use scheduler config from state store to enable/disable preemption
2018-10-30 11:06:32 -05:00
1805032e69
Fix linting and better comments
2018-10-30 11:06:32 -05:00
cc295b90de
Implement preemption for system jobs.
...
This commit implements an allocation selection algorithm for finding
allocations to preempt. It currently special cases network resource asks
from others (cpu/memory/disk/iops).
2018-10-30 11:06:32 -05:00
22aee7294e
Merge branch 'f-fix-resource-type' of github.com:hashicorp/nomad into f-fix-resource-type
2018-10-16 18:30:12 -05:00
53c3f8151b
fix linting
2018-10-16 18:29:49 -05:00
a78cefec18
use int64
2018-10-16 15:34:32 -07:00
7c0d8c646c
Change CPU/Disk/MemoryMB to int everywhere in new resource structs
2018-10-16 16:21:42 -05:00
f5a76d8411
review comments
2018-10-15 15:31:13 -07:00
7ecd65109a
Check constraints on devices
2018-10-14 13:35:47 -07:00
5284554fcc
rework device checker
2018-10-13 16:47:53 -07:00
1089e13b14
add to stack
2018-10-13 12:27:49 -07:00
9b5aaac410
Device feasability checker
2018-10-13 12:27:49 -07:00
1574e898af
Fix bug in reconciler where terminal allocs on a job already stopped were unnecessarily updated
2018-10-08 21:03:49 -05:00
01f8e5b95f
renames
2018-10-04 14:57:25 -07:00
52f9cd7637
fixing tests
2018-10-04 14:26:19 -07:00
bac5cb1e8b
Scheduler uses allocated resources
2018-10-02 17:08:25 -07:00
a10118c461
Add failed follow up to the list of allowed eval trigger reasons
...
needs unit test
2018-09-25 10:49:55 -07:00
6a21f9fe96
Unique TriggerBy for blocked evals
...
Give blocked evals a unique triggerby reason to make debugging a chain
of evaluations easier.
2018-09-24 14:47:49 -07:00
3c19d01d7a
server
2018-09-15 16:23:13 -07:00
3ba62efd5e
Failed/paused deployments do not block migrations
...
This PR changes behavior of the scheduler such that a task group with a
deployment that is failed or paused will not cause the scheduler to skip
migrations.
The reason for this change is that it causes a bad UX when draining
nodes with allocations that are part of a failed/paused deployment.
These operations should not be coupled in any way and this remedies
that.
Prior behavior was still correct, but required either jobs to
transistion to a healthy state or for the node to hit its drain
deadline.
2018-09-10 15:28:45 -07:00
cc92cd92cd
Merge pull request #4642 from hashicorp/b-vet
...
Fix vet errors and use newer go version in travis
2018-09-04 17:04:02 -07:00
c6576ddac1
Fix make check errors
2018-09-04 16:03:52 -07:00
751c0eb5a5
code review feedback
2018-09-04 16:10:11 -05:00
9bc0962527
Track top k nodes by norm score rather than top k nodes per scorer
2018-09-04 16:10:11 -05:00
6ed527c636
Use heap to store top K scoring nodes.
...
Scoring metadata is now aggregated by scorer type to make it easier
to parse when reading it in the CLI.
2018-09-04 16:10:11 -05:00
65cf4373b3
fix linting error
2018-09-04 16:10:11 -05:00
dd5fe6373f
Fix scoring logic for uneven spread to incorporate current alloc count
...
Also addressed other small code review comments
2018-09-04 16:10:11 -05:00
e72c0fe527
more cleanup
2018-09-04 16:10:11 -05:00
4c624424e6
added some unit tests for -1 spread score
2018-09-04 16:10:11 -05:00
92d37acc2a
comment and formatting cleanup
2018-09-04 16:10:11 -05:00
7b0a27cad6
fix scoring algorithm when min count == current count
2018-09-04 16:10:11 -05:00
bad075f640
Remove hardcoded boosts for even spread.
...
instead, calculate them based on delta between current and minimum value
2018-09-04 16:10:11 -05:00
c56873ff37
Implement support for even spread across datacenters, with unit test
2018-09-04 16:10:11 -05:00
d091c00dd3
Support implicit spread target to account for remaining desired counts
2018-09-04 16:10:11 -05:00
33779abe5f
fix comments
2018-09-04 16:10:11 -05:00
5812f906c8
Allow empty spread targets, and validate target percentages.
2018-09-04 16:10:11 -05:00
55f276c189
Include spreads configured at job level when precomputing weights/desired counts.
2018-09-04 16:10:11 -05:00
fbd0004707
Fix warnings
2018-09-04 16:10:11 -05:00
db0d95b09c
Implement spread iterator that scores according to percentage of desired count in each target.
...
Added this as a new step in the stack and some unit tests
2018-09-04 16:10:11 -05:00
eccf128c5c
Some minor changes from code review
2018-09-04 16:10:11 -05:00
038ed52877
Fix after rename to ConstraintSetContainsAny
2018-09-04 16:10:11 -05:00
3a39db3902
Fix linting
2018-09-04 16:10:11 -05:00
d5cd2bbddb
Remove unnecessary reset
2018-09-04 16:10:11 -05:00
dccb693221
test for setcontainsany, and treat set_contains same as set_contains_all
2018-09-04 16:10:11 -05:00
70bfd0c0cb
Address some review feedback
2018-09-04 16:10:11 -05:00
8685593ec0
Back out changes to propertyset that were not necessary for affinities
2018-09-04 16:10:11 -05:00
5eacd6ada4
Implement affinity support in generic scheduler
2018-09-04 16:10:11 -05:00
e1c239daae
Merge pull request #4414 from hashicorp/b-stop-summary
...
Reset Queued allocs to zero when job stopped
2018-07-16 14:32:55 -07:00
6b6777359b
scheduler: fix missing err assignment
2018-07-11 14:27:10 -04:00
5f6def5b04
scheduler: better error handling
2018-07-05 11:00:03 -04:00
030e650e78
scheduler: fix nil pointer exception
2018-07-02 16:05:38 -04:00
300b1a7a15
Tests only use testlog package logger
2018-06-13 15:40:56 -07:00
c3c79c408e
Reset Queued allocs to zero when job stopped
...
When a job is stopped but not purged, we should set the Queued count to
be zero.
2018-06-13 10:46:39 -07:00
b64788043e
make test create index clearer
2018-06-05 17:29:59 -05:00
3e264dcb79
Fix reconciler bug with deployment not being created if job create index is different
...
This fixes an issue where if a job is purged and resubmitted Nomad does not create
a new deployment. Adds unit test that failed before this fix
2018-06-05 13:58:53 -05:00
f8a23bc54a
fix test comment
2018-05-09 16:01:34 -05:00
ef531b0f34
Add unit tests for forced rescheduling
2018-05-09 11:30:42 -05:00
c1b92c284e
Work in progress - force rescheduling of failed allocs
2018-05-08 17:26:57 -05:00
555d14fd92
Add test
2018-05-07 14:55:01 -05:00
cf44670d56
Make sure that task group has a deployment state before using it
2018-05-07 14:55:01 -05:00
c6478d9469
clarify comment
2018-05-07 14:55:01 -05:00
768fec8505
Allow healthy canary deployment to skip progress deadline
2018-05-07 14:55:01 -05:00
8626c1b94a
Reschedule when we have canaries properly
2018-05-07 14:55:01 -05:00
8dee3ab068
canary reschedule test
2018-05-07 14:50:01 -05:00
deb93dc7b7
Test for rescheduling when there are canaries
2018-05-07 14:50:01 -05:00
550f5e31f8
Allow canary count greater than desired
2018-05-07 14:50:01 -05:00
f95ab4ade8
Mark canaries on creation, and unmark on promotion
2018-05-07 14:50:01 -05:00
5329900f6d
Only use DesiredTransition.Reschedule in reconciler when its an active deployment
2018-05-07 14:50:01 -05:00
e7444c3873
Add test where deployment is marked as complete when done even with failed allocs
2018-05-07 14:50:01 -05:00
57969b4ee0
fix reconcile tests
2018-05-07 14:50:01 -05:00
5547974f35
Only reschedule allowed deployment allocs
2018-05-07 14:50:01 -05:00
fcf4f582d0
small review feedback fixes
2018-05-07 14:50:01 -05:00
1336002255
Progress deadline in deployment state
2018-05-07 14:50:01 -05:00
ee50789c22
Initial implementation
2018-05-07 14:50:01 -05:00
a569d34f25
Add custom status description for rescheduling follow up evals, and make unit test robust
2018-04-10 15:30:15 -05:00
e5b5803265
Only mark allocs as part of deployment if deployment is active
2018-04-05 15:40:49 -07:00
7e17bc231f
remove unnecessary check and other fixes from code review
2018-04-04 07:35:20 -05:00
00537c739b
Fixes edge cases around timing and task finish time being set more than once
2018-04-03 16:34:59 -05:00
3aa4ee9d75
Fix lost handling of not actually down nodes
2018-03-30 14:17:41 -07:00
d87e528059
rename skip->ignore and improve comment formatting
2018-03-29 15:11:10 -05:00
38a7614776
Refactored for readability, pair programmed with @dadgar
2018-03-29 13:28:37 -05:00
5090fefe96
Filter out allocs with DesiredState = stop, and unit tests
2018-03-29 09:28:52 -05:00
b18f789020
Unmark drain when nodes hit their deadline and only batch/system left and add all job type integration test
2018-03-28 17:25:58 -07:00
d2899728fd
Fix linting
2018-03-28 12:26:28 -05:00
9d60e2cebf
Correct status desc on draining system allocs
2018-03-26 17:54:46 -07:00
33e170c15d
s/linear/constant/g
2018-03-26 14:45:09 -05:00
5668c3c38e
Merge pull request #4037 from hashicorp/b-fix-terminal-filtering-service-allocs
...
Fix edge case in reconciler
2018-03-26 13:14:51 -05:00
1b9e413a1a
one field per line in struct definition
2018-03-26 13:13:21 -05:00
e106da84de
name and test
2018-03-26 11:06:21 -07:00
e2a6e64fca
Don't create unnecessary deployments
2018-03-23 16:55:21 -07:00
cbfd69ce7a
Fix edge case in reconciler where service jobs with ClientstatusComplete were not replaced
2018-03-23 18:41:00 -05:00
3b72dd94ba
Do not mark an allocation as an inplace update if specification hasn't changed
2018-03-23 14:36:05 -07:00
cb61a4bdc7
Fix linting errors
2018-03-21 16:51:45 -07:00
92b636dd32
Fix deadline handling
2018-03-21 16:51:44 -07:00
9263cc2ed7
scheduler: migrate non-terminal migrating allocs
...
filterByTainted node should always migrate non-terminal migrating allocs
2018-03-21 16:49:48 -07:00
d1ec65d765
switch to new raft DesiredTransition message
2018-03-21 16:49:48 -07:00
db4a634072
RPC, FSM, State Store for marking DesiredTransistion
...
fix build tag
2018-03-21 16:49:48 -07:00
c0542474db
drain: initial drainv2 structs and impl
2018-03-21 16:49:48 -07:00
329605b7cc
fix up scheduling test
2018-03-21 15:54:03 -04:00
60f12d206f
improve comments; update watchDriver
2018-03-21 15:15:26 -04:00
d92703617c
simplify logic
...
bump log level
2018-03-21 15:15:26 -04:00
d8f68e5ef8
fix up codereview feedback
2018-03-21 15:15:26 -04:00
c7fd0bd8a1
fix up scheduler mocks
2018-03-21 15:15:25 -04:00
3aa726baab
fix scheduler driver name; create node structs file
2018-03-21 15:15:25 -04:00
3cba95e8a7
allow nomad to schedule based on the status of a client driver health check
...
Slight updates for go style
2018-03-21 15:15:25 -04:00
56e60e5840
Fix linting warning
2018-03-14 16:12:22 -05:00
9a5e6edf1f
Rename DelayCeiling to MaxDelay
2018-03-14 16:10:32 -05:00
3e96c6c4e0
Address more code review feedback
2018-03-14 16:10:32 -05:00
9fed0d2103
Get reschedule policy from the alloc directly
2018-03-14 16:10:32 -05:00
e89bbf7289
Update comment about WaitTime
2018-03-14 16:10:32 -05:00
e2656ef546
Cleaner handling of batched evals
2018-03-14 16:10:32 -05:00
47e0280d96
More small review feedback
2018-03-14 16:10:32 -05:00
2ba976dec8
Remove unnecessary check against 5 second window for determining immediate scheduling eligibility
2018-03-14 16:10:32 -05:00
5373ade731
Scheduler and Reconciler changes to support delayed rescheduling
2018-03-14 16:10:32 -05:00
e0f6a33fe5
spelling: system
2018-03-11 19:01:19 +00:00
a89e1b8395
spelling: strategy
2018-03-11 18:58:19 +00:00
f8eb766fb5
spelling: reschedulable
2018-03-11 18:48:12 +00:00
ed8db9992e
spelling: feasibility
2018-03-11 18:07:09 +00:00
bf9283c606
spelling: corresponding
2018-03-11 17:51:41 +00:00
ca4ceb0e5c
spelling: commits
2018-03-11 17:47:45 +00:00
7b6ba7a1f4
Fixes bug in reconciler where previously rescheduled allocs are rescheduled again. Simplified logic and added test case to catch this.
2018-02-20 12:07:56 -06:00
7c57303dd2
Clarify comment
2018-02-05 16:37:07 -06:00
d48c411692
Reconciler should consider failed allocs when marking deployment as failed.
2018-02-02 19:40:25 -06:00
a1237d627a
code review feedback
2018-01-31 09:58:05 -06:00
5ad892026a
Add a field to track the next allocation during a replacement
2018-01-31 09:58:05 -06:00
2ed4de7e7b
Track previous node id correctly, plus unit test
2018-01-31 09:58:05 -06:00
dd4917c2f0
Add more clarification in comment
2018-01-31 09:58:05 -06:00
09bef7d1ce
Preallocate slice for skipped nodes
2018-01-31 09:58:05 -06:00
237beb49ae
Better score threshold
2018-01-31 09:58:05 -06:00
fa18c0def4
Add one more unit test
2018-01-31 09:58:05 -06:00
a75540cec6
Limit iterator uses a score threshold and a maxSkip value to be able to skip lower scoring nodes
2018-01-31 09:58:05 -06:00
b6268a5fab
Beef up unit test for rescheduling batch jobs
2018-01-31 09:56:53 -06:00
ea4a889e28
Address more code review feedback
2018-01-31 09:56:53 -06:00
bd89d2b39e
Make sure that reschedule trackers are not added for node drain replacements
2018-01-31 09:56:53 -06:00
a662b38801
Improve reconciler unit tests
2018-01-31 09:56:53 -06:00
fee4ccf154
Prevent side effect modification of select options when preferred nodes are set
2018-01-31 09:56:53 -06:00
21b7b79d5d
Add helper methods, use require and other code review feedback
2018-01-31 09:56:53 -06:00
d0f9d59abb
Reconile with changes to structs for reschedule tracking
2018-01-31 09:56:53 -06:00
fbb1936dee
Fix some comments and lint warnings, remove unused method
2018-01-31 09:56:53 -06:00
031c566ada
Reschedule previous allocs and track their reschedule attempts
2018-01-31 09:56:53 -06:00
fd2fbefa4c
Add a field to track the next allocation during a replacement
2018-01-24 17:55:05 -06:00
6dda0ebaed
gofmt
2018-01-04 14:45:15 -08:00
2f561609b7
Fix detection of successful batch allocations
...
This PR restores older behavior of detecting successful batch
allocations (04d86ffd1006fde9dfb2ca8c1237fe60b995b0e3). This has the
side effect that we correctly filter desired status stop but not
successful batch allocations and create their replacements.
2018-01-04 14:20:32 -08:00
1712b03705
Merge branch 'master' into 0.8
2018-01-03 16:06:38 -06:00
51bd0b59c7
Return an error if evaluation doesn't exist in state store at plan apply time.
2017-12-18 14:55:36 -06:00
3c36abfe14
Update eval modify index as part of plan apply.
2017-12-18 10:03:55 -06:00
3b4d7ac2a3
Fix some typos
2017-12-14 13:29:27 -06:00
45494f7304
Fix port labels on mock Alloc/Job/Node
2017-12-08 14:50:06 -08:00
44240ce440
Merge pull request #3375 from hashicorp/b-batch
...
Allow batch jobs to be rerun if purged
2017-10-13 17:11:45 -07:00
c1cc51dbee
sync
2017-10-13 14:36:02 -07:00
746cd7403f
Allow batch jobs to be rerun if purged
...
This PR allows batch jobs to be rerun if they have been purged.
2017-10-13 12:40:37 -07:00
a66c53d45a
Remove `structs` import from `api`
...
Goes a step further and removes structs import from api's tests as well
by moving GenerateUUID to its own package.
2017-09-29 10:36:08 -07:00
4173834231
Enable more linters
2017-09-26 15:26:33 -07:00
3904bde9a3
Fix batch handling of complete allocs/node drains
...
This PR fixes:
* An issue in which a node-drain that contains a complete batch alloc
would cause a replacement
* An issue in which allocations with the same name during a scale
down/stop event wouldn't be properly stopped.
* An issue in which batch allocations from previous job versions may not
have been stopped properly.
Fixes https://github.com/hashicorp/nomad/issues/3210
2017-09-14 15:08:57 -07:00
84d06f6abe
Sync namespace changes
2017-09-07 17:04:21 -07:00
0aef02a4f9
fix test
2017-08-21 14:07:54 -07:00
27256ebcc6
Placing allocs counts towards placement limit
...
This PR makes placing new allocations count towards the limit. We do not
restrict how many new placements are made by the limit but we still
count towards the limit. This has the nice affect that if you have a
group with count = 5 and max_parallel = 1 but only 3 allocs exist for it
and a change is made, you will create 2 more at the new version but not
destroy one, taking you down to two running as you would have
previously.
Fixes https://github.com/hashicorp/nomad/issues/3053
2017-08-21 12:41:19 -07:00
2453f13fc5
fixes
2017-08-15 12:27:05 -07:00
0570e09feb
Fix panic occuring from improper bitmap size
...
This PR fixes an allignment calculation when determining the bitmap
size.
Fixes https://github.com/hashicorp/nomad/issues/3008
2017-08-12 15:37:02 -07:00
f0ced87b95
fixed all spelling mistakes for goreport
2017-08-07 17:13:05 -04:00
7b13c0d702
Lost allocs replaced even if deployment failed
...
This PR allows the scheduler to replace lost allocations even if the job
has a failed or paused deployment. The prior behavior was confusing to
users.
Fixes https://github.com/hashicorp/nomad/issues/2958
2017-08-03 17:42:14 -07:00
7d2b84ab01
Review fixes
2017-08-01 14:18:52 -07:00
2650bb1d12
Distinct Property supports arbitrary limit
...
This PR enhances the distinct_property constraint such that a limit can
be specified in the RTarget/value parameter. This allows constraints
such as:
```
constraint {
distinct_property = "${meta.rack}"
value = "2"
}
```
This restricts any given rack from running more than 2 allocations from
the task group.
Fixes https://github.com/hashicorp/nomad/issues/1146
2017-07-31 16:52:13 -07:00
4f69355a66
Fix incorrect destructive update with distinct_property constraint
...
This PR fixes an issue in which an update to a task group with a
distinct property constraint would result in an incorrect destructive
update.
2017-07-31 11:17:35 -07:00
5f1f91a46c
Use go-testing-interface instead of testing
...
This drops the testings stdlib pkg from our dependencies. Saves a
whopping 46kb on our binary (was really hoping for more of a win there),
but also avoids potential ugliness with how testing sets flags.
2017-07-25 15:35:19 -07:00
492239d3ee
Improve multiple group handling in a deployment
...
This PR resolves a bug in which a job with multiple task groups would
create new deployment objects each, thus clearing out all other task
groups deployment state.
2017-07-25 11:27:47 -07:00
184bfd4836
Better comment
2017-07-20 12:31:08 -07:00
248315a2d9
Handle destructive changes before placements
...
This PR updates the generic scheduler to handle destructive changes
before handling placements. This is important because the destructive
change may be due to a lowering of resources. If this is the case, the
handling of the destructive changes first may make it possible for the
placement to happen.
To reason about this imagine there is one node with CPU = 500.
If the group originally had:
* `count = 1`
* `cpu = 400`
And then the job was updated such that the group had:
* `count = 4`
* `cpu = 120`
If the original alloc isn't discounted first, nothing would be able to
place.
2017-07-20 12:24:27 -07:00
ce265e0aff
Update full node test to test more advanced case
2017-07-20 12:23:40 -07:00
a9ec1d6ca7
Fix update limit calculation to avoid panic
...
This PR fixes the rolling update limit calculation to avoid a panic when
there are more allocations for a deployment that haven't determined
their health than the max_parallel count of the task group.
Fixes https://github.com/hashicorp/nomad/issues/2820
2017-07-19 11:11:47 -07:00
22e84d00ab
Fix deep copy of driver config
2017-07-17 17:53:21 -07:00
641e178416
Stop before trying to place
2017-07-17 17:18:12 -07:00
66a90326e1
Treat destructive updates atomically
2017-07-16 10:35:38 -07:00
f86760db3c
Basic logs
2017-07-07 16:49:08 -07:00
Alex Dadgar