Examples: query, "exact match", wildcard*, wild?ard, wild*rd
Fuzzy search: cake~ (finds cakes, bake)
Term boost: "red velvet"^4, chocolate^2
Field grouping: tags:(+work -"fun-stuff")
Escaping: Escape characters +-&|!(){}[]^"~*?:\ with \, e.g. \+
Range search: properties.timestamp:[1587729413488 TO *] (inclusive), properties.title:{A TO Z}(excluding A and Z)
Combinations: chocolate AND vanilla, chocolate OR vanilla, (chocolate OR vanilla) NOT "vanilla pudding"
Field search: properties.title:"The Title" AND text
Answered
Hi

Hi TimelyPenguin76 AgitatedDove14 SuccessfulKoala55 ,does a task support to be set custom parameters? I want to use k8s.py(/glue) to create a Job in Kubernetes, which need some custom labels on it. There labels are important to our platform which will be integrated with ClearML. If it is impossible, is there any solution to solve this kind of problem? Thanks in advance!

  
  
Posted 3 years ago
Votes Newest

Answers 3


Well, you can always change the k8s glue code to be able to set pod labels based on task properties. If that's a valuable addition and works for you, we would love a PR 🙂

  
  
Posted 3 years ago

Hi FloppyDeer99 ,
When using the ClearML Agent K8S glue, you provide a pod template - in that template, under the metadata.labels section you can specify any number of custom labels you'd like to set on the pods started by the K8s glue

  
  
Posted 3 years ago

SuccessfulKoala55 I know it. Here is an example. I have a queue named q1 and two groups, which are used to manage the quota of our platform. In this situation, all tasks which may belong to group1 or group2 will be enqueued into the q1. And ClearML Agent K8S glue will watch the q1 and pull tasks from it. However, K8S does not know the group which this task uses except this information can be get from task. I can solve this problem by creating queues for every group. But there are too many groups (maybe over 50), I think this is not a good idea. So is there any solution which may be more elegant?

  
  
Posted 3 years ago