Permalink
Commits on Sep 16, 2010
  1. bug 2236 - don't printf unused dbpath anymore

    the local sqlite/dbpath feature of libconflate was confusing,
    so don't allude to it anymore in the -vvv output.
    
    Change-Id: I849f72128b4859cf69f0b1a1a3cde6a93a99211c
    Reviewed-on: http://review.northscale.com/2304
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 7, 2010
  2. usage typo and width fix

    Change-Id: Ia77b17f9c9614434234bd4f4f5162416c46bddce
    Reviewed-on: http://review.northscale.com/2492
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 16, 2010
Commits on Sep 15, 2010
  1. mcmux-prefixed "A:HOST:PORT version" command

    With tweaks so that it's not a broadcast command.
    
    Change-Id: Id3941efa4255d43cd012000c6be149bcf2a07f19
    Reviewed-on: http://review.northscale.com/2437
    Reviewed-by: Manik Taneja <mtaneja@zynga.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    mtaneja committed with steveyen Sep 15, 2010
  2. bug 2345 - don't crash when some binary clients send in empty keys

    One client that does this is dbmaint, so you can test this fix by
    pointing dbmaint against moxi, and moxi doesn't crash anymore...
    
      ./ep-engine/management/dbmaint --vacuum --port=11211
    
    Change-Id: Ie1908333cb16d3b7a2f33c85ce5e34694db39de9
    Reviewed-on: http://review.northscale.com/2427
    Reviewed-by: Matt Ingenthron <matt@northscale.com>
    Reviewed-by: Trond Norbye <trond.norbye@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 15, 2010
  3. Use the default automake rule to detect endian

    Change-Id: Ic09e0f2d3db93fdfd6d7d861476588ebacfc9a13
    Reviewed-on: http://review.northscale.com/2367
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    trondn committed with steveyen Sep 10, 2010
  4. Skip check for malloc and realloc

    Change-Id: I21ed07ed90e1ccc6fffd6b1336fb0ab6d37df7ba
    Reviewed-on: http://review.northscale.com/2366
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    trondn committed with steveyen Sep 10, 2010
  5. mcmux compat needs cmd-line behavior

    Change-Id: I2ced858d2e00e75a89f189c2b5cb9a4e8fa2059f
    Reviewed-on: http://review.northscale.com/2426
    Reviewed-by: Manik Taneja <mtaneja@zynga.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 15, 2010
Commits on Sep 14, 2010
  1. bug 2337, moved MOXI_BLOCKING_CONNECT definition

    Moved to a place where windows can find it, as windows builder doesn't
    appear to use config_static.h.
    
    Change-Id: Ibca5f1b07b025ed58945a1959648324febbe67b5
    Reviewed-on: http://review.northscale.com/2419
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 14, 2010
  2. misc changes for mcmux compatiblity

    1. version, quit and stats proxy are be treated as self commands
       and do not require a prefix
    2. gets should return SERVER_ERROR for a non existent server when
       running in mcmux compatiblity mode
    
    Change-Id: I3b9c231001955d80f4a8274c2f884c31cec2fe2e
    Reviewed-on: http://review.northscale.com/2417
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    mtaneja committed with steveyen Sep 14, 2010
  3. bug 2337, default to blocking connect() on windows platform

    Change-Id: Ic370eef54da59732aa38463c3de9cce48dab169b
    Reviewed-on: http://review.northscale.com/2412
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 14, 2010
  4. stats for connect error cases

    Change-Id: I3d499d69425a6693511e4e0eec8debe5c72f4292
    Reviewed-on: http://review.northscale.com/2403
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 14, 2010
  5. check for downstream_conn_max to limit number of connections

    If moxi reaches a max number of downstream connections (per host_ident
    per thread), the zstored_acquire_downstream_conn() function will
    return that condition (via an extra out param in its API).
    
    After that, the downstream has to wait until there's an available
    downstream_conn, which it does using the new
    zstored_downstream_waiting_add() function (and all the related,
    supporting downstream_waiting_head/tail linked-list/queue management
    code).
    
    As a downstream conn is released, moxi assigns the downstream conn to
    a waiting downstream.
    
    When all downstream conns are closed (for a given host_ident &
    thread), moxi processes any waiting downstreams.
    
    Change-Id: I52e754069a959f319aca545f7de33470892681cc
    Reviewed-on: http://review.northscale.com/2402
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 13, 2010
  6. downstream_waiting_head/tail started

    Change-Id: I061361b34774ec4493383eb9e00a1b988b65e0a2
    Reviewed-on: http://review.northscale.com/2401
    Reviewed-by: Jayesh Jose <jjose@zynga.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 13, 2010
  7. added downstream_conn_max config parsing

    Change-Id: Icef738e732eae3b1205bcdfb50d59b8da6e4c2f6
    Reviewed-on: http://review.northscale.com/2400
    Reviewed-by: Jayesh Jose <jjose@zynga.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 13, 2010
Commits on Sep 13, 2010
  1. windows needs EINPROGRESS, EISCONN

    Change-Id: I5295bd136a9b8cf65d5ef0feaaa54dfccec81cb4
    Reviewed-on: http://review.northscale.com/2312
    Reviewed-by: Trond Norbye <trond.norbye@gmail.com>
    Reviewed-by: Matt Ingenthron <matt@northscale.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 8, 2010
  2. bug 2237 - calling zstored_error_count() more during errors

    A zstored_error_count() invocation was incorrectly being guarded by an
    IF-THEN statement.  So, this fix moves the call out of the IF body,
    but doing so meant having to change zstored_error_count()'s function
    signature (to just take a simple host_ident string).
    
    Also, refactored out a zstored_get_downstream_conns() helper function,
    to handle a case when the zstored_error_count() wasn't counting
    correctly the very first time before moxi had tried its very first
    connection attempt.
    
    Change-Id: I08468db8d767ea555c893d6c39276dc93968b92e
    Reviewed-on: http://review.northscale.com/2378
    Reviewed-by: Matt Ingenthron <matt@northscale.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 13, 2010
  3. connect_max_errors/retry config in stats proxy

    Change-Id: Id1b5df1fb44079fafa1ae67b5beea56840be71cc
    Reviewed-on: http://review.northscale.com/2377
    Reviewed-by: Matt Ingenthron <matt@northscale.com>
    Tested-by: Matt Ingenthron <matt@northscale.com>
    steveyen committed with Matt Ingenthron Sep 13, 2010
Commits on Sep 10, 2010
  1. Add .libs to list of ignored files

    Change-Id: I7137473043bcad450d6347b5f6d114beb611799f
    Reviewed-on: http://review.northscale.com/2365
    Reviewed-by: Dustin Sallings <dustin@spy.net>
    Tested-by: Dustin Sallings <dustin@spy.net>
    trondn committed with dustin Sep 10, 2010
Commits on Sep 9, 2010
  1. conn might be NULL during debug logging

    ...especially for a broadcast command (like stats or flush_all).
    
    Change-Id: I493af84c96bc1e5098b52dc336281aa22a2e32c9
    Reviewed-on: http://review.northscale.com/2302
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 7, 2010
  2. from zstored, 5 sec timeout during non-blocking connect()

    As part of this change, refactored out a update_event_timed() helper
    function.
    
    Change-Id: Idae1834c21cfb35e1030bc1e3f33b9d65a1be46a
    Reviewed-on: http://review.northscale.com/2301
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 7, 2010
Commits on Sep 8, 2010
  1. fix support for unix domain sockets

    Change-Id: Id9efc18e2e5a658bf198cd37b5852a11f1f2a1ef
    Reviewed-on: http://review.northscale.com/2320
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    mtaneja committed with steveyen Sep 8, 2010
  2. mcmux protocol support for moxi

    Added support for ASCII protocol extension to moxi. To enable
    mcmux compatiblity -X option should be passed as one of the command
    line paramters along with the port number that moxi should listen on.
    
    ./moxi -p 11311 -X
    
    when started in this mode, moxi expects the mcmux protocol string
    to be appended before the start of a memcache command.
    An example of an interaction between client and moxi :
    
    A:localhost:11411 get k1
    VALUE k1 0 5
    manik
    END
    
    get k1
    ERROR
    
    Amended version: check for null d->upstream_conn
    Use a stack variable for msst instead of allocating it.
    
    Change-Id: I40d853023c80d08a40113e338e5bb6708e7f4bd9
    Reviewed-on: http://review.northscale.com/2317
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    mtaneja committed with steveyen Sep 8, 2010
  3. connect timeout and retry interval are now runtime configurable

    Usage...
    
      ./moxi -Z connect_max_errors=10,connect_retry_interval=10000,other_flags=here
    
    Change-Id: I8de9e94e9f6e23fa9a8214add476af699c423616
    Reviewed-on: http://review.northscale.com/2289
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 6, 2010
  4. checking MAX_DOWNSTREAM_CONNECTION_ERRORS and DOWNSTREAM_RETRY_INTERVAL

    Change-Id: I2c3574f5564a45cdf916c66261759d6d2276b605
    Reviewed-on: http://review.northscale.com/2288
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 6, 2010
  5. fixed connection leak

    Change-Id: I6542dc05d1569365961e872fe9420dc7eeca6244
    Reviewed-on: http://review.northscale.com/2283
    Reviewed-by: Manik Taneja <mtaneja@zynga.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 6, 2010
  6. tracking connect_failed stat

    Change-Id: Ib783cae1283845909a473a2df08e4a6643037f9a
    Reviewed-on: http://review.northscale.com/2280
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 5, 2010
  7. more assert()

    Change-Id: Ib073e89cd54ca3b5fca8a5fd0631969d3953154b
    Reviewed-on: http://review.northscale.com/2279
    Reviewed-by: Dustin Sallings <dustin@spy.net>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 5, 2010
  8. removed unused downstream_waiting/next_waiting

    Change-Id: I453b6865463035934e364dc74eeb69919ef1a5d2
    Reviewed-on: http://review.northscale.com/2278
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 5, 2010
  9. more stats around acquire/release

    Change-Id: Iafec2e02fd4ee665845db7c8ad153d83bff09377
    Reviewed-on: http://review.northscale.com/2277
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 5, 2010
  10. dead code removal

    Change-Id: I3bb4bb55ba0af75116a68a2106a23883435a10bc
    Reviewed-on: http://review.northscale.com/2271
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 3, 2010
  11. downstream_conn->extra now set/cleared during acquire/release

    The association between a downstream connection and a downstream
    connection is now correctly set or cleared during
    zstored_acquire/release_downstream_conn()
    
    Change-Id: Ifa1687f629fed8882a3f07f9f21f6230aa724be4
    Reviewed-on: http://review.northscale.com/2270
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 3, 2010
  12. updated TODO comment

    Change-Id: I4b1266b788687af6a64fa3cf80756eee7c7573f9
    Reviewed-on: http://review.northscale.com/2266
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 3, 2010
  13. allow conn_connect callback to control drive_machine() loop

    This is useful when a newly connected downstream conn is getting
    assigned to a downstream struct, and we don't actually want
    drive_machine() to go through the conn_pause logic for the newly
    connected downstream conn.
    
    Change-Id: I46d5cf00294c95392c865e7753583bfb8fd290dc
    Reviewed-on: http://review.northscale.com/2265
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 3, 2010
  14. invoking downstream_connect_init() after non-blocking connects()

    Change-Id: Iaa7d361f92a177cca4e7e4d902130744f72da572
    Reviewed-on: http://review.northscale.com/2264
    Tested-by: Steve Yen <steve.yen@gmail.com>
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 3, 2010
  15. cproxy_forward_or_error

    Change-Id: I72add23d34758bf7dd163e2d478ac16eb7faefbe
    Reviewed-on: http://review.northscale.com/2263
    Reviewed-by: Steve Yen <steve.yen@gmail.com>
    Tested-by: Steve Yen <steve.yen@gmail.com>
    steveyen committed Sep 3, 2010