!SDKYnTgYyOXNBpJySp:matrix.org

cylc protocols & APIs

21 Members
HTTPS, WebSocket, ZeroMQ; REST, GraphQL1 Servers

Load older messages


SenderMessageTime
21 Sep 2022
@dwsutherland:matrix.orgDavid Sutherland
In reply to @oliver-sanders2:matrix.org
Interestingly I haven't seen an intercycle dependency edge the wrong way around yet.

I'm guessing there's a single interface in the code somewhere which has the pair the wrong way around. I might have messed it up with the UID work.
*

Fixed:
https://github.com/cylc/cylc-flow/pull/5149

it was a simple wrong way round.. However only for edges created by following the parent path (upstream), hence why you never saw intercycle dependencies the wrong way (since they are probably defined forward in time usually (always?), and the edge already exists from following the child path)

05:12:56
@dwsutherland:matrix.orgDavid Sutherland *

Fixed:
https://github.com/cylc/cylc-flow/pull/5149

it was a simple wrong way round.. However only for edges created by following the parent path (upstream), hence why you never saw intercycle dependencies the wrong way (since they are probably defined forward in time usually (always?), and the edge already exists from following the child (downstream) path)

05:13:36
@revilo666:matrix.orgHilary OliverNice 🎉05:16:12
@oliver-sanders2:matrix.orgoliver sandersBrill, cheers!08:32:40
22 Sep 2022
@dwsutherland:matrix.orgDavid Sutherland

Been trying to get an idea of the job for updating graphene/graphql-core ... and whether to strip it back so we can avoid graphgl-ws (and even graphql-tornado) dependency, or whether we try to support those projects ourselves..

The new graphql-core/graphene versions don't use those rx observables ... graphql-ws is quite small, and relies on these..

I didn't want to build more stuff relying on the old versions, but because of the potentially large time sink involved .. I think I'll have to build the cylc-flow graphql subscriptions based on what we are currently using .. then move on to the upgrade

03:18:29
27 Sep 2022
@dwsutherland:matrix.orgDavid Sutherlandgraphene-tornado has had a contribution to make it work with the latest graphene (and a beta release for it).. a quite small change 👍21:12:37
@dwsutherland:matrix.orgDavid Sutherland * Been trying to get an idea of the job for updating graphene/graphql-core ... and whether to strip it back so we can avoid graphgl-ws (and even graphene-tornado) dependency, or whether we try to support those projects ourselves..
The new graphql-core/graphene versions don't use those rx observables ... graphql-ws is quite small, and relies on these..
I didn't want to build more stuff relying on the old versions, but because of the potentially large time sink involved .. I think I'll have to build the cylc-flow graphql subscriptions based on what we are currently using .. then move on to the upgrade
21:13:02
@revilo666:matrix.orgHilary OliverThat's good news!21:25:00
28 Sep 2022
@oliver-sanders2:matrix.orgoliver sandersGood news indeed!08:40:12
@oliver-sanders2:matrix.orgoliver sanders With that released it's only graphql-ws that's holding us back from graphene 3. Unfortunately that repo remains static, might be worth poking them about this, there are a few PRs kicking about, namely this one which is important for us. 08:42:57
20 Oct 2022
@dwsutherland:matrix.orgDavid Sutherland

The UIS stop extension:
https://github.com/cylc/cylc-uiserver/blob/da038f00a1c9d6fb6bbf5b2a156bbc4fc7b299a7/cylc/uiserver/app.py#L516-L524

Runs when stopping (keyboard interrupt) the hubless server (cylc gui), however, it isn't run when stopping the server via the hub..

From it's contents I assume it should be run with both forms of stop

03:59:11
@revilo666:matrix.orgHilary OliverHmm yeah looks like it.04:00:37
@dwsutherland:matrix.orgDavid Sutherlandfound this out while trying to clean up api tokens .. the non-hub stop cleans them up04:01:04
@revilo666:matrix.orgHilary OliverRedacted or Malformed Event04:01:42
@dwsutherland:matrix.orgDavid Sutherlandwill make and issue for it .. had a search, can't figure it out yet.. there are hooks .. but the hub should be able (?) to use the same stop extension, surely .. maybe it's brute killing it (or not probably shutting it down)04:02:51
@dwsutherland:matrix.orgDavid Sutherlandhttps://github.com/cylc/cylc-uiserver/issues/39204:38:05
17 Nov 2022
@revilo666:matrix.orgHilary Oliver Small datastore issue noticed while testing the graph view: a single old succeeded task got stuck in the datastore (showed up in all views, and cylc dump, so not a graph issue). I think it probably resulted from changing the n-window extent, which I was playing around with, but I'm not sure. 02:28:47
@dwsutherland:matrix.orgDavid Sutherlanddata-store or pool.. The pruning essentially takes the set difference of what’s in the active pool and root/active store and uses this as the basis for pruning.. If I can reproduce it then I’ll fix02:52:20
@dwsutherland:matrix.orgDavid Sutherland* data-store or pool.. The pruning essentially takes the set difference of what’s in the active pool and root/active/n=0 store and uses this as the basis for pruning.. If I can reproduce it then I’ll fix02:53:41
@dwsutherland:matrix.orgDavid Sutherland * data-store or pool.. The pruning essentially takes the set difference of what’s in the active pool and root/active/n=0 store and uses this as the basis for pruning (those branch/n>0 nodes recorded-against/associated-with these node(s)).. If I can reproduce it then I’ll fix 02:56:22
@revilo666:matrix.orgHilary OliverPretty sure it wouldn't be the scheduler task pool, but I'll try to reproduce it with task pool logging turned on. Don't waste any time looking at the moment.03:14:18
@metronnie:matrix.orgRonnie DuttaAt one point I saw an old succeeded task showing up as running (or possibly submitted) in the tree view but not graph or table views. It didn't have a job icon.10:02:05
@oliver-sanders2:matrix.orgoliver sanders metronnie: I think that a submitted task which was a UI datastore issue, now fixed. 10:03:10
@oliver-sanders2:matrix.orgoliver sanders * metronnie: I think that was a submitted task which was a UI datastore issue, now fixed. 10:03:18
22 Nov 2022
@dwsutherland:matrix.orgDavid Sutherland

Put up and issue for CLI via Hub/proxy and/or UIS:
https://github.com/cylc/cylc-flow/issues/5235

Easiest part done (making the API info available), need to create a client and config entry for http on the cylc-flow end now

00:50:10
@dwsutherland:matrix.orgDavid Sutherland *

Put up and issue for CLI via Hub/proxy and/or UIS:
https://github.com/cylc/cylc-flow/issues/5235

Still some peripheral work to do, but, first part done (making the API info available)..
Now need to create a client and config entry for http on the cylc-flow end now. Initially queries/mutations only, but will aim to have websockets working (which should give Oliver something to play with on the TUI front, until we have GraphQL subscriptions at the Scheduler)

00:54:20
@dwsutherland:matrix.orgDavid Sutherland *

Put up and issue for CLI via Hub/proxy and/or UIS:
https://github.com/cylc/cylc-flow/issues/5235

Still some peripheral work to do, but, first part done (making the API info available)..
Now need to create a client and config entry for http on the cylc-flow end. Initially queries/mutations only, but will aim to have websockets working (which should give Oliver something to play with on the TUI front, until we have GraphQL subscriptions at the Scheduler)

01:14:07
7 Dec 2022
@dwsutherland:matrix.orgDavid Sutherland

going via the UIS:

(uiserver) sutherlander@graphic-vbox:~$ cylc ping --debug linear
2022-12-07T21:18:59+13:00 DEBUG - Loading site/user config
    files
2022-12-07T21:18:59+13:00 DEBUG - Reading file
    /home/sutherlander/.cylc/flow/8.1.0.dev/global.cylc
2022-12-07T21:18:59+13:00 DEBUG - http:send {'command':
    'graphql', 'args': {'request_string': '\nquery ($wFlows: [ID]) {\n
    workflows(ids: $wFlows) {\n    id\n    name\n    port\n    pubPort\n
    }\n}\n', 'variables': {'wFlows': ['linear/run1']}}, 'meta': {'prog': 'ping',
    'host': 'graphic-vbox', 'comms_method': 'http'}}
2022-12-07T21:18:59+13:00 DEBUG - http:recv {'data':
    {'workflows': [{'id': '~sutherlander/linear/run1', 'name': 'linear/run1',
    'port': 43072, 'pubPort': 43070}]}}
linear/run1 running on graphic-vbox:43072 43070
08:19:41
@dwsutherland:matrix.orgDavid Sutherland is there a general communication method global config setting? all I could find is the platform one, which as far as I can tell is only for jobs:
https://cylc.github.io/cylc-doc/nightly/html/reference/config/global.html#global.cylc[platforms][%3Cplatform%20name%3E]communication%20method
08:22:39
@dwsutherland:matrix.orgDavid Sutherland Will have to create one otherwise.. had to $ export CYLC_TASK_COMMS_METHOD='http' 08:23:37

Show newer messages


Back to Room ListRoom Version: 1