#hypothes.is

/

      • NOTICE: [h] tilgovi force-pushed auth-isolation from 8d3a813 to 9b908fd: https://github.com/hypothesis/h/commits/auth-isolation
      • NOTICE: h/auth-isolation 9b908fd Randall Leeds: Tighten up the identity and auth interaction...
      • GitHub168 has left the channel
      • nickstenn
        tilgovi: i read the article on auth/authz/identity in pyramid -- really interesting
      • tilgovi
        seemed smart to me
      • a few nice insights
      • I should annotate it
      • "knob on a knob"
      • nickstenn
        i think i might need to reread it a few times...
      • scharf joined the channel
      • tilgovi
        nickstenn: it's wonderful how it well it actually maps to what i'm doing on the frontend right now
      • i was pleasantly reminded when I brought it up for you
      • scharf has quit
      • Treora joined the channel
      • nickstenn
        tilgovi: Treora and I are discussing this reindexing stuff
      • I agree with your latest comments
      • except for the recommendation that it go into run.py
      • tilgovi
        the most recent suggestion with run.py?
      • no?
      • okay, a different script
      • nickstenn
        run.py is supposed to be an example application, not the canonical way of running the thing
      • tilgovi
        makes sense to use the .cfg, though?
      • nickstenn
        so yes, use the .cfg for the current source
      • and a reindexer where you simply specify the destination and (optionally) an alias
      • tilgovi
        old new --alias would have been okay, too
      • nickstenn
        python reindex.py annotator-2014-10-03 --alias annotator
      • tilgovi
        but i think I like this
      • nickstenn
        tbh I'm okay with either
      • Treora
        by the way, I think I did not use the .cfg because it's quite flask-specific.
      • tilgovi
        m
      • so you're right
      • Treora
        but other than that it is indeed desirable that the information is read from the config file
      • tilgovi
        although it does appear to just parse as python
      • so execfile for instance
      • but
      • ugh
      • nickstenn
        okay, compromise
      • sod the config file
      • tilgovi
        lol
      • old new --alias?
      • nickstenn
        python reindex.py old new [--alias aliasname]
      • tilgovi
        yeah
      • nickstenn
        so very close to store#103
      • vannevar
        Simple alternative reindexing tool: https://github.com/openannotation/annotator-sto...
      • nickstenn
        but make the reindexer a script
      • tilgovi
        thanks for your patience with this, Treora
      • I know I'm insane about this stuff
      • Treora
        thanks for yours
      • nickstenn
        if you get a chance to do that, Treora, fab
      • if not, I'll do it later
      • Treora
        nickstenn: isn't that closer to #93?
      • nickstenn
        store#93
      • vannevar
        Create reindex.py, add warning on mapping conflict: https://github.com/openannotation/annotator-sto...
      • Treora
        just the command line syntax is different: $ python reindex.py --reindex old new --alias new old
      • nickstenn
        yes, you're right
      • but the CLI can be simplified substantially
      • make --host the option
      • tilgovi
        yeah
      • oh
      • nickstenn
        and new and old are positionals
      • Treora
        though this way you can nicely separate the steps, do an alias without doing reindex for example
      • nickstenn
        Treora: if you're that user, you can just use curl :)
      • this is about providing an easy path for people who don't understand ElasticSearch, right?
      • tilgovi
        yah, this is the essential point
      • Treora
        yes
      • tilgovi
        it's the balance that I felt was off
      • trying to do too much
      • when at that point it's hard to decide what's best for the developer anymore
      • because trade-offs and things
      • Treora
        indeed
      • nickstenn
        the steady state, when 'annotator' is already an alias to a dated index, should look like:
      • python reindex.py --host localhost:9200 --alias annotator annotator annotator-$(date +"%Y-%m-%d")
      • tilgovi
        uh
      • you have an extra "annotator" after --alias or no?
      • nickstenn
        tilgovi: that's the <from>
      • tilgovi
        oh
      • i see
      • my brain inserted an old/new before --alias
      • Treora
        from annotator to annotator?
      • tilgovi
        nah, that's from annotator to annotator-blahblah
      • with the --alias annotator
      • nickstenn
        you reindex from an alias to a dated index, and then update the alias to point to the (new) dated index
      • Treora
        yes ok this behaviour is good, there is just the issue when annotator was not an alias, but a real index
      • nickstenn
        Treora: that's fine
      • tilgovi
        then you don't specify --alias
      • nickstenn
        you blow up
      • tilgovi
        or you learn :-D
      • nickstenn
        you can't create an alias on top of an index, and we can't make the tool pretend that we can :)
      • that's why I did try and clarify that that was the "steady state" usage
      • tilgovi
        q how do you feel about running reindex before and after the alias update?
      • fine if you'd rather not
      • nickstenn
        no, don't
      • Treora
        why twice?
      • tilgovi
        writes can happen between reindex finishing and the alias command
      • doing it twice catches those
      • it's what i always do
      • nickstenn
        Treora: because the (reindex, update alias) sequence isn't atomic
      • but again
      • tilgovi
        and neither are writes to es
      • nickstenn
        this is for joe average user
      • tilgovi
        yah
      • nickstenn
        who doesn't have traffic so wild that they have this problem
      • tilgovi
        :-P
      • so, us :-D
      • nickstenn
        we can certainly document the limitations of the tool
      • but KISS/YAGNI, etc.
      • tilgovi
        yepyepyep
      • nickstenn
        Treora: have we confused you even more or is that clear now?
      • Treora
        I have all the time tried to also make things work when we're not in steady state, that's why I came up with more complex scripts
      • GitHub193 joined the channel
      • NOTICE: [h] tilgovi force-pushed auth-isolation from 9b908fd to 7e8f786: https://github.com/hypothesis/h/commits/auth-isolation
      • NOTICE: h/auth-isolation 7e8f786 Randall Leeds: Tighten up the identity and auth interaction...
      • GitHub193 has left the channel
      • so one can delete the old index, and then create an alias
      • nickstenn
        Treora: it still does work
      • python reindex.py annotator annotator-$date
      • and update your config file
      • Treora
        yes that's a workaround
      • nickstenn
        if you want to do fancy shit with indices, you should probably understand how indices work first
      • Treora
        ok so then that workaround is acceptable.
      • that simplifies the whole thing a lot
      • nickstenn
        definitely acceptable
      • firstdate joined the channel
      • scharf joined the channel
      • scharf has quit
      • Treora
        nickstenn: #93 is using your suggested command line syntax now
      • nickstenn
        awesome, thank you Treora
      • Treora
        I need to run
      • Perhaps some of the documentation of #103 could still be useful? If it's okay I'll leave the finishing touch to you
      • cheers
      • Treora has quit
      • savesoff_ has quit
      • savesoffomak joined the channel
      • Sebastien-L has quit
      • savesoffomak has quit
      • scharf joined the channel
      • firstdate has quit
      • scharf has quit
      • csillag joined the channel
      • GitHub50 joined the channel
      • NOTICE: [h] tilgovi force-pushed auth-isolation from 7e8f786 to eacede6: https://github.com/hypothesis/h/commits/auth-isolation
      • NOTICE: h/auth-isolation b601178 Randall Leeds: Tighten up the identity and auth interaction...
      • NOTICE: h/auth-isolation 50a3dd1 Randall Leeds: Test the events emitted by the auth controller
      • NOTICE: h/auth-isolation 3658025 Randall Leeds: Move session service back to h module...
      • GitHub50 has left the channel
      • Sebastien-L joined the channel
      • nickstenn
        tilgovi: I recently heard about this thing called sleep
      • it's pretty cool
      • tilgovi
        yeah, i'm doing it now
      • nickstenn
        :)