Two-phase HDB commit via iprop log, + GC for log

We used to update the iprop log and HDB in different orders depending on
the kadm5 operation, which then led to various race conditions.

The iprop log now functions as a two-phase commit (with roll forward)
log for HDB changes.  The log is auto-truncated, keeping the latest
entries that fit in a configurable maximum number of bytes (defaults to
50MB).  See the log-max-size parameter description in krb5.conf(5).

The iprop log format and the protocol remain backwards-compatible with
earlier versions of Heimdal.  This is NOT a flag-day; there is NO need
to update all the slaves at once with the master, though it is advisable
in general.  Rolling upgrades and downgrades should work.

The sequence of updates is now (with HDB and log open and locked):

a) check that the HDB operation will succeed if attempted,
b) append to iprop log and fsync() it,
c) write to HDB (which should fsync()),
d) mark last log record committed (no fsync in this case).

Every kadm5 write operation recover transactions not yet confirmed as
committed, thus there can be at most one unconfirmed commit on a master
KDC.

Reads via kadm5_get_principal() also attempt to lock the log, and if
successful, recover unconfirmed transactions; readers must have write
access and must win any race to lock the iprop log.

The ipropd-master daemon also attempts to recover unconfirmed
transactions when idle.

The log now starts with a nop record whose payload records the offset of
the logical end of the log: the end of the last confirmed committed
transaction.  This is kown as the "uber record".  Its purpose is
two-fold: act as the confirmation of committed transactions, and provide
an O(1) method of finding the end of the log (i.e., without having to
traverse the entire log front to back).

Two-phase commit makes all kadm5 writes single-operation atomic
transactions (though some kadm5 operations, such as renames of
principals, and changes to principals' aliases, use multiple low-level
HDB write operations, but still all in one transaction).  One can still
hold a lock on the HDB across many operations (e.g., by using the lock
command in a kadmin -l or calling kadm5_lock()) in order to push
multiple transactions in sequence, but this sequence will not be atomic
if the process or host crashes in the middle.

As before, HDB writes which do not go through the kadm5 API are excluded
from all of this, but there should be no such writes.

Lastly, the iprop-log(1) command is enhanced as follows:

 - The dump, last-version, truncate, and replay sub-commands now have an
   option to not lock the log.  This is useful for inspecting a running
   system's log file, especially on slave KDCs.

 - The dump, last-version, truncate, and replay sub-commands now take an
   optional iprop log file positional argument, so that they may be used
   to inspect log files other than the running system's
   configured/default log file.

Extensive code review and some re-writing for clarity by Viktor Dukhovni.
This commit is contained in:
Nicolas Williams
2015-05-29 15:53:40 -05:00
parent d774aeda38
commit 20df2c8706
33 changed files with 3274 additions and 804 deletions

View File

@@ -38,7 +38,7 @@
.Os Heimdal
.Sh NAME
.Nm iprop-log
.Nd maintain the iprop log file
.Nd examine and maintain the iprop log file
.Sh SYNOPSIS
.Nm
.Op Fl Fl version
@@ -54,7 +54,20 @@
.Fl Fl realm= Ns Ar string
.Xc
.Oc
.Oo Fl K Ar integer \*(Ba Xo
.Fl Fl keep-entries= Ns Ar integer
.Xc
.Oc
.Oo Fl B Ar integer \*(Ba Xo
.Fl Fl max-bytes= Ns Ar integer
.Xc
.Oc
.Oo Fl R \*(Ba Xo
.Fl Fl reset
.Xc
.Oc
.Op Fl h | Fl Fl help
.Op Ar log-file
.Pp
.Nm iprop-log dump
.Oo Fl c Ar file \*(Ba Xo
@@ -65,7 +78,16 @@
.Fl Fl realm= Ns Ar string
.Xc
.Oc
.Oo Fl n \*(Ba Xo
.Fl Fl no-lock
.Xc
.Oc
.Oo Fl R \*(Ba Xo
.Fl Fl reverse
.Xc
.Oc
.Op Fl h | Fl Fl help
.Op Ar log-file
.Pp
.Nm iprop-log replay
.Op Fl Fl start-version= Ns Ar version-number
@@ -79,6 +101,35 @@
.Xc
.Oc
.Op Fl h | Fl Fl help
.Op Ar log-file
.Pp
.Nm iprop-log last-version
.Oo Fl c Ar file \*(Ba Xo
.Fl Fl config-file= Ns Ar file
.Xc
.Oc
.Oo Fl r Ar string \*(Ba Xo
.Fl Fl realm= Ns Ar string
.Xc
.Oc
.Oo Fl n \*(Ba Xo
.Fl Fl no-lock
.Xc
.Oc
.Op Fl h | Fl Fl help
.Op Ar log-files
.Pp
.Nm iprop-log signal
.Oo Fl c Ar file \*(Ba Xo
.Fl Fl config-file= Ns Ar file
.Xc
.Oc
.Oo Fl r Ar string \*(Ba Xo
.Fl Fl realm= Ns Ar string
.Xc
.Oc
.Op Fl h | Fl Fl help
.Pp
.Sh DESCRIPTION
Supported options:
.Bl -tag -width Ds
@@ -94,11 +145,42 @@ command can be one of the following:
configuration file
.It Fl r Ar string , Fl Fl realm= Ns Ar string
realm
.It Xo
.Fl K Ar integer ,
.Fl Fl keep-entries= Ns Ar integer
.Xc
.It Xo
.Fl B Ar integer ,
.Fl Fl max-bytes= Ns Ar integer
.Xc
.It Xo
.Fl R ,
.Fl Fl reset
.Xc
.El
.Pp
Truncates the log. Sets the new logs version number for the to the
last entry of the old log. If the log is truncted by emptying the
file, the log will start over at the first version (0).
If
.Fl Fl reset
is given, then the given, configured, or default log file will be
truncated and will start at version 1. This forces full propagations to
all slave KDCs.
.Pp
Otherwise the log will be truncated but some entries will be preserved,
as specified by the
.Fl Fl keep-entries
and/or
.Fl Fl max-bytes
options. The largest number of
.Fl Fl keep-entries
entries that are available and fit in the given
.Fl Fl max-bytes
option will be used. The
.Fl Fl keep-entries option defaults to 100, and the
.Fl Fl max-bytes
option defaults to the
.Ar log-max-size
parameter in the configuration.
.Pp
.It dump
.Bl -tag -width Ds
.It Fl c Ar file , Fl Fl config-file= Ns Ar file
@@ -108,9 +190,23 @@ configuration file
.Fl Fl realm= Ns Ar string
.Xc
realm
.It Xo
.Fl n Ar string ,
.Fl Fl no-lock
.Xc
.It Xo
.Fl R Ar string ,
.Fl Fl reverse
.Xc
.El
.Pp
Print out all entries in the log to standard output.
Print out all entries in the given, configured, or default log file to
standard output. If the
.Fl n
option is used then don't lock the iprop log file. If the
.Fl R
option is used, then print the entries in reverse order
(this can be useful when the log is very large).
.It replay
.Bl -tag -width Ds
.It Fl Fl start-version= Ns Ar version-number
@@ -125,17 +221,34 @@ configuration file
realm
.El
.Pp
Replay the changes from specified entries (or all if none is
specified) in the transaction log to the database.
Replay the changes from specified entries (or all if none is specified)
in the given, configured, or default transaction log file to the
database.
.It last-version
.Bl -tag -width Ds
.It Fl c Ar file , Fl Fl config-file= Ns Ar file
configuration file
.It Fl r Ar string , Fl Fl realm= Ns Ar string
realm
.It Xo
.Fl n Ar string ,
.Fl Fl no-lock
.Xc
.El
.Pp
prints the version of the last record in each of the given log files, or
the configured, or the default log file if none is given.
.It signal
.Bl -tag -width Ds
.It Fl c Ar file , Fl Fl config-file= Ns Ar file
configuration file
.It Fl r Ar string , Fl Fl realm= Ns Ar string
realm
.El
.Pp
prints the version of the last log entry.
Signals the ipropd-master daemon to send updates to slaves. Normally
kadmin does this every time it writes to the database, so this should
rarely be needed.
.El
.Sh SEE ALSO
.Xr iprop 8