Switch branches/tags
Commits on Dec 20, 2013
Commits on Nov 11, 2013
  1. Removig (now) useless ugprade code

    ferd committed Nov 11, 2013
  2. Add better resizing

    ferd committed Nov 11, 2013
    Always resize *up* after a connection iff the log buffer is smaller than
    the current default value for the buffer.
  3. More flexible TCP Syslog drain shrinking

    ferd committed Nov 11, 2013
    Shrinking is now dependent on number of messages in the buffer, presence
    of dropped messages, and number of sequential failures instead of time
Commits on Nov 6, 2013
  1. Upgrade script

    ferd committed Nov 6, 2013
  2. Fixing the buffer resizing

    ferd committed Nov 6, 2013
Commits on Oct 31, 2013
  1. Longer timeouts on live upgrades

    ferd committed Oct 31, 2013
Commits on Oct 30, 2013
  1. v69.8->v69.9 live upgrade code

    ferd committed Oct 30, 2013
  2. Merge pull request #62 from heroku/dynamic_buf_size

    ferd committed Oct 30, 2013
    Dynamic buf size
Commits on Oct 29, 2013
  1. shrinking of HTTP drains

    ferd committed Oct 29, 2013
    Inactive drains (disconnected for too long) get shrunk down, and resized
    when going back online.
Commits on Oct 25, 2013
  1. Force shrinking on never-connected drains

    ferd committed Oct 25, 2013
    Only a successful connection will grow the drain size.
  2. Add tests for TCP drain resizing

    ferd committed Oct 25, 2013
  3. Fix buffer resizing bug

    ferd committed Oct 25, 2013
    When a buffer gets resized and needs to drop data, the data is dropped,
    but the maximal value of the buffer is never brought down. This patch
    makes it so the function works as expected.
    This fix also removes all the unused conversion functions that were used
    in a past deploy.
Commits on Oct 24, 2013
  1. Dynamic resize

    ferd committed Oct 24, 2013
    If a buffer is inactive for too long, its size gets shrunk to 10 instead
    of 1024, and gets re-established when reconnecting.
Commits on Oct 23, 2013
  1. upgrade procedure for v69.8

    ferd committed Oct 23, 2013
Commits on Oct 21, 2013
  1. Merge branch 'v69.7'

    ferd committed Oct 21, 2013
  2. Merge pull request #61 from heroku/disable-api

    archaelus committed Oct 21, 2013
    Adding read-only & disabled mode for Logplex API
  3. Adding read-only & disabled mode for Logplex API

    ferd committed Oct 21, 2013
    The Disabled mode entirely shuts down access to the web API at all
    levels, including for healthchecks. Use with extreme care.
    The read-only mode forbids all non-GET operations outside of what v2
    canaries are allowed to do.
Commits on Oct 8, 2013
  1. Merge pull request #57 from heroku/no_tail_warning

    archaelus committed Oct 8, 2013
    Adds L20 - tail session forbidden message.
Commits on Sep 24, 2013
  1. Making logplex go on R16B02

    ferd committed Sep 24, 2013
Commits on Sep 23, 2013
  1. Upgrade path for v69.5->v69.6

    ferd committed Sep 23, 2013
Commits on Sep 19, 2013
  1. Merge pull request #58 from heroku/update-l2met-format

    ferd committed Sep 19, 2013
    stop using obsolete l2met format
  2. Merge pull request #55 from heroku/m_equal_logplex_stats

    ferd committed Sep 19, 2013
    Add m= in front of logplex_stats logging so it's easier to group in splunk
Commits on Sep 18, 2013
Commits on Sep 16, 2013
Commits on Sep 5, 2013
  1. Changing memory allocation settings for binaries

    ferd committed Sep 5, 2013
    Changing the allocator strategy for binary_alloc to do aobf rather than
    bf (+MBas aobf). Depending on how all of the binaries are allocated,
    this could make new allocations favor the same carrier. This will only
    add a small cpu overhead when allocating new binaries. This could
    however worsen our utilization even more if we are unlucky, it will
    however make allocations faster and more in the same area in the
    expected case, which should reduce fragmentation and untraceable leaks.
    We're also decreasing the size of our mbcs. Right now we have smbcs set
    to 256 kb and lmbcs at 5 MB (rounded up to 8MB as ERTS only allocates
    multiples of 2) and an average multi-block carrier size of 7.78 MB. We
    try to set +MBlmbcs 512 so that we get many more carriers and
    thus increase the chance that it can be returned to the OS.
    These two options have been recommended by members of the Erlang/OTP
    team to reduce the passive memory leaks due to allocation patterns
    compared to our peculiar use cases for log messages.
Commits on Sep 3, 2013
  1. Basic batchio inclusion, replaces lager macros

    ferd committed Sep 3, 2013
    batches io:format calls into a buffer process to reduce the number
    of calls required and also do overload protection automatically.
    Experimental material to see if it helps with performance, given the
    buffering and load shedding allow to go asynchronous on log messages
    without loosing too much data.