Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Scott Kurz
    @scottkurz
    camel-case slightly less reflective of a "hierarchy" than boost.liberty.xyz, boost.liberty.security.xyz could be but not sure how much that matters or how far to take that
    YK Chang
    @yeekangc
    Scott, can you elaborate on how you think it will be like application.properties?
    I can see Boost supporting a set of general properties and there will still be runtime-specific properties (defined under their respective namespace).
    Scott Kurz
    @scottkurz
    Well.. if Boost is generating all or most of the structured server.xml config then all you have left to configure is a set of properties... which could be supplied in a number of ways, including application.properties.
    (And I appreciate moving a bit slowly/carefully for moving in yet another config direction.. after all I'd think we want to support use cases like using MPConfig APIs to access these properties.. and how many passes do you need to do for one variable substitution referencing another.)
    Andrew Guibert
    @aguibert
    would it be possible for boost to generate server.xml config into the src/ dir rather than build output?
    The elephant in the room w.r.t. boost generating config is:
    A) how does the user customize it
    B) how does it fit into containers
    I think that generating some/all of the config into src/ could alleviate both of those
    Scott Kurz
    @scottkurz
    generating into src sounds like a step in the wrong direction to me... A) you can customize using configDropins to some degree...hopefully over time the need to do so will lessen. B) not sure how generating config into src helps
    Andrew Guibert
    @aguibert
    for (B) its easier to do COPY src/main/liberty/config /config/ than to reach into the build folder, where there may be dirty artifacts like logs, workarea, etc that you dont' want to scoop up
    Scott Kurz
    @scottkurz
    there wouldn't be dirty stuff in the target config though
    Andrew Guibert
    @aguibert
    depends when the docker build runs, I assume it runs at/near the end?
    Scott Kurz
    @scottkurz
    yeah I was assuming it would... (the current boost goal isnt' doing anything BTW)
    Andrew Guibert
    @aguibert
    so if we start the server for integration tests and such it'll have logs and potentially a loose app xml in it
    Scott Kurz
    @scottkurz
    Well.. not to say we've completely thought this through.. but not sure why we wouldn't copy just exactly what we need plus the whole config dir.
    Andrew Guibert
    @aguibert
    @scottkurz when we launch ITs with Boost, do you think Boost has any opportunity to set sysprops on the JVM that gets forked for ITs?
    Andrew Guibert
    @aguibert
    Scott Kurz
    @scottkurz
    I think we need to expose the libertyRuntimeVersion prop support : MicroShed/boost#360
    Scott Kurz
    @scottkurz
    And no.. I didn't think Boost would be setting sysprops on the IT JVM.. But MST is setting such props right? Like microshed.hostname ?
    If that's the case I'd wonder if MST's property setting should incorporate any other boost-related props/values
    Andrew Guibert
    @aguibert
    do we thinkg boost#360 can be done in time for OC1?
    that's a must IMO, or at least update to 19008 as the new baseline
    Scott Kurz
    @scottkurz
    yeah we should at least hard code 19008 by Tuesday or so
    (if we don't get further)
    Andrew Guibert
    @aguibert
    so for Boost+MST interop we need to do 1 of 2 things:
    A) have Boost NOT start/stop the server for the integration-test phase
    B) have Boost set some sys props on the spawned IT JVM
    actually, does (A) even happen OOTB?
    Scott Kurz
    @scottkurz
    no A) doesn't happen OOB
    Andrew Guibert
    @aguibert
    ok, so in theory it should be possible for some tool besides Boost to start the runtime, right?
    however, this seems tricky because Boost generates the server.xml and knows what props need to be set
    Scott Kurz
    @scottkurz
    Boost's package has already generated config as defaultVals of server vars
    YK Chang
    @yeekangc
    0.1 of the Liberty Dev VS Code extension is now in the marketplace: https://marketplace.visualstudio.com/items?itemName=Open-Liberty.liberty-dev-vscode-ext
    YK Chang
    @yeekangc
    Here’s a summary of what’s new/changed in 3.0 of Liberty Maven plugin: https://github.com/OpenLiberty/ci.maven/releases/tag/liberty-maven-3.0/
    Scott Kurz
    @scottkurz
    So in testing OpenLiberty/ci.maven#580 I asked the question "how to do a local .m2 install of a nightly build?" given that we don't have maven snapshots. It seems like an answer is maybe a short command like: mvn install:install-file -Dfile=ol.zip -DgroupId=io.openliberty -DartifactId=openliberty-runtime -Dversion=19.0.0.10-SNAPSHOT -Dpackaging=zip after downloading something like this to ol.zip.
    This seems like it can be pretty useful. Is this worth documenting somewhere? I'm guessing it'd break if we needed to use the function to go retrieve features....but we don't since the features are all there?
    YK Chang
    @yeekangc
    May be somewhere under open-liberty and/or ci.maven, @scottkurz?
    Lee Surprenant
    @lmsurpre
    Not sure if this is the best place to ask this or not, but here it goes: we have a multi-module Maven project that builds a war file and we notice that, when we import these projects into eclipse (with Liberty Tools), it generates a corresponding EAR project for every dependency of the war project.
    the accepted answer claims this is an artifact of the WebSphere tools, but I havn't had a chance to try this in an eclipse installation that doesn't have liberty tools yet
    unfortunately, we can't seem to find a way to disable/workaround this annoying behavior. we can either:
    1. disable wtp activation as described in that stack overflow post, then document how to enable project facets for the 2 web projects in our list; or
    2. instruct users to manually delete all the generated EAR projects
    Lee Surprenant
    @lmsurpre
    in the case of number 1, i found that even if i just manually enable the Dyanmic Web Module facet, clicking "Apply and Close" will generate a corresponding EAR project (but at least it doesn't generate an EAR project for all of its dependencies as well)
    any hints/tips for us?
    Lee Surprenant
    @lmsurpre
    OK, I confirmed that I only see this behavior with the Liberty Tools installed. Where is the most appropriate place to open an issue?
    YK Chang
    @yeekangc
    Hold on. Let me grab you the link.
    Lee Surprenant
    @lmsurpre
    thank you YK. i have the 2019-06 Liberty Tools installed now...i'm updating to 2019-09 to confirm its still an issue and then will open there
    YK Chang
    @yeekangc
    Thank you, Lee. Please file an issue when ready.
    Scott Kurz
    @scottkurz
    Do we have any examples on conditionally installing and activating a feature via LMP? It's easy enough to guard the install via a profile... but what's a good way to conditionally write a configDropins activating it... on top of the rest of the config (server.xml, etc.) ?
    Scott Kurz
    @scottkurz
    Maybe what I'm asking is do you have to get into copying/deleting the file or is there some combo of existing goals/config that does this already?
    YK Chang
    @yeekangc
    What do you mean by “conditionally”, Scott? install-feature would install features for you if they are declared in server configuration but not present yet.
    Scott Kurz
    @scottkurz
    I mean like "guarded" by a profile so it only happens if the profile is active
    so yeah... if you had a nice conditional way to set up the server config you wouldn't need to configure, say, the "install-feature" goal based on profile... you'd already have solved the problem.
    I'm just saying I could see how to do something like configure "install-feature" but what's a good way to do this for server.xml?