diff options
Diffstat (limited to '')
| -rw-r--r-- | proto/LMDB_README.html | 421 |
1 files changed, 421 insertions, 0 deletions
diff --git a/proto/LMDB_README.html b/proto/LMDB_README.html new file mode 100644 index 0000000..a9794cc --- /dev/null +++ b/proto/LMDB_README.html @@ -0,0 +1,421 @@ +<!doctype html public "-//W3C//DTD HTML 4.01 Transitional//EN" + "http://www.w3.org/TR/html4/loose.dtd"> + +<html> + +<head> + +<title>Postfix OpenLDAP LMDB Howto</title> + +<meta http-equiv="Content-Type" content="text/html; charset=utf-8"> + +</head> + +<body> + +<h1><img src="postfix-logo.jpg" width="203" height="98" ALT="">Postfix OpenLDAP LMDB Howto</h1> + +<hr> + +<h2>Introduction</h2> + +<p> Postfix uses databases of various kinds to store and look up +information. Postfix databases are specified as "type:name". OpenLDAP +LMDB (called "LMDB" from here on) implements the Postfix database +type "lmdb". The name of a Postfix LMDB database is the name of +the database file without the ".lmdb" suffix. </p> + +<p> This document describes: </p> + +<ul> + +<li> <p> <a href="#with_lmdb">Building Postfix with LMDB support</a>. +</p> + +<li> <p> <a href="#configure">Configuring LMDB settings</a>. </p> + +<li> <p> <a href="#locking">Using LMDB maps with non-Postfix programs</a>. </p> + +<li> <p> <a href="#supported"> Required minimum LMDB patchlevel</a>. </p> + +<li> <p> <a href="#credits"> Credits</a>. </p> + +</ul> + +<h2><a name="with_lmdb">Building Postfix with LMDB support</a></h2> + +<p> Postfix normally does not enable LMDB support. To +build Postfix with LMDB support, use something like: </p> + +<blockquote> +<pre> +% make makefiles CCARGS="-DHAS_LMDB -I/usr/local/include" \ + AUXLIBS_LMDB="-L/usr/local/lib -llmdb" +% make +</pre> +</blockquote> + +<p> If your LMDB shared library is in a directory that the RUN-TIME +linker does not know about, add a "-Wl,-R,/path/to/directory" option after +"-llmdb". </p> + +<p> Postfix versions before 3.0 use AUXLIBS instead of AUXLIBS_LMDB. +With Postfix 3.0 and later, the old AUXLIBS variable still supports +building a statically-loaded LMDB database client, but only the new +AUXLIBS_LMDB variable supports building a dynamically-loaded or +statically-loaded LMDB database client. </p> + +<blockquote> + +<p> Failure to use the AUXLIBS_LMDB variable will defeat the purpose +of dynamic database client loading. Every Postfix executable file +will have LMDB database library dependencies. And that was exactly +what dynamic database client loading was meant to avoid. </p> + +</blockquote> + + +<p> Solaris may need this: </p> + +<blockquote> +<pre> +% make makefiles CCARGS="-DHAS_LMDB -I/usr/local/include" \ + AUXLIBS_LMDB="-R/usr/local/lib -L/usr/local/lib -llmdb" +% make +</pre> +</blockquote> + +<p> The exact pathnames depend on how LMDB was installed. </p> + +<p> When building Postfix fails with: </p> + +<blockquote> +<pre> +undefined reference to `pthread_mutexattr_destroy' +undefined reference to `pthread_mutexattr_init' +undefined reference to `pthread_mutex_lock' +</pre> +</blockquote> + +<p> Add the "-lpthread" library to the "make makefiles" command. </p> + +<blockquote> +<pre> +% make makefiles .... AUXLIBS_LMDB="... -lpthread" +</pre> +</blockquote> + +<h2><a name="configure">Configuring LMDB settings</a></h2> + +<p> Postfix provides one configuration parameter that controls +LMDB database behavior. </p> + +<ul> + +<li> <p> lmdb_map_size (default: 16777216). This setting specifies +the initial LMDB database size limit in bytes. Each time a database +becomes "full", its size limit is doubled. The maximum size is the +largest signed integer value of "long". </p> + +</ul> + +<h2> <a name="locking">Using LMDB maps with non-Postfix programs</a> </h2> + +<p> Programs that use LMDB's built-in locking protocol will corrupt +a Postfix LMDB database or will read garbage. </p> + +<p> Postfix does not use LMDB's built-in locking protocol, because +that would require world-writable lockfiles, and would violate +Postfix security policy. Instead, Postfix uses external locks based +on fcntl(2) to prevent writers from corrupting the database, and +to prevent readers from receiving garbage. </p> + +<p> See lmdb_table(5) for a detailed description of the locking +protocol that all programs must use when they access a Postfix LMDB +database. </p> + +<h2> <a name="supported"> Required minimum LMDB patchlevel </a> </h2> + +<p> Currently, Postfix requires LMDB 0.9.11 or later. The required +minimum LMDB patchlevel has evolved over time, as the result of +Postfix deployment experience: </p> + +<ul> + +<li> <p> LMDB 0.9.11 allows Postfix daemons to log an LMDB error +message, instead of falling out of the sky without any notification. +</p> + +<li> <p> LMDB 0.9.10 closes an information leak where LMDB was +writing up to 4-kbyte chunks of uninitialized heap memory to the +database. This would persist information that was not meant to be +persisted, or share information that was not meant to be shared. +</p> + +<li> <p> LMDB 0.9.9 allows Postfix to use external (fcntl()-based) +locks, instead of having to use world-writable LMDB lock files, +violating the Postfix security model in multiple ways. </p> + +<li> <p> LMDB 0.9.8 allows Postfix to recover from a "database full" +error without having to close the database. This version adds support +to update the database size limit on-the-fly. This is necessary +because Postfix database sizes vary with mail server load. </p> + +<li> <p> LMDB 0.9.7 allows the postmap(1) and postalias(1) commands +to use a bulk-mode transaction larger than the amount of physical +memory. This is necessary because LMDB supports databases larger +than physical memory. </p> + +</ul> + +<h2> <a name="credits"> Credits</a> </h2> + +<ul> + +<li> <p> Howard Chu contributed the initial Postfix dict_lmdb driver. +</p> + +<li> <p> Wietse Venema wrote an abstraction layer (slmdb) that +behaves more like Berkeley DB, NDBM, etc. This layer automatically +retries an LMDB request when a database needs to be resized, or +after a database was resized by a different process. </p> + +<li> <p> Howard and Wietse went through many iterations with changes +to both LMDB and Postfix, with input from Viktor Dukhovni. </p> + +</ul> + +<!-- + +<h2><a name="limitations">Unexpected failure modes of Postfix LMDB +databases. </a> </h2> + +<p> As documented below, conversion to LMDB introduces a number of +failure modes that don't exist with other Postfix databases. Some +failure modes have been eliminated in the course of time. +The writeup below reflects the status as of LMDB 0.9.9. </p> + +--> + +<!-- + +<p> <strong>Unexpected "Permission denied" errors. </strong></p> + +<dl> + +<dt> Problem: </dt> <dd> <p> A world-readable LMDB database cannot +be opened by a process with a UID that differs from the database +file owner, even when an attempt is made to open the database +read-only. This problem does not exist with other Postfix databases. +</p> </dd> + +<dt> Background: </dt> <dd> <p> The LMDB implementation requires +write access to maintain read locks, and perhaps for other purposes. +</p> </dd> + +<dt> Solution: </dt> <dd> <p> Consider using cdb: to manage root-owned +databases under the root-owned <tt>/etc</tt> or config_directory +(default: <tt>/etc/postfix</tt>) such as access(5), virtual(5), +transport(5). Support to create LMDB databases is available only +for unprivileged Postfix daemon processes such as postscreen(8), +tlsmgr(8) and verify(8) that manage postfix-owned databases under +the postfix-owned data_directory (default: <tt>/var/lib/postfix</tt>). +</p> </dd> + +</dl> + +--> + +<!-- + +<p> <strong>Unexpected "readers full" errors. </strong></p> + +<dl> + +<dt> Problem: </dt> <dd> <p> Under heavy load, database read +operations fail with MDB_READERS_FULL errors. This problem does not +exist with other Postfix databases. </p> </dd> + +<dt> Background: </dt> <dd> <p> The LMDB implementation enforces a +hard limit on the number of simultaneous read requests for the same +database environment. This limit must be specified in advance with +the lmdb_max_readers configuration parameter. </p> </dd> + +<dt> Mitigation: </dt> <dd> <p> Postfix logs a warning suggesting +that the lmdb_max_readers parameter value be increased, and retries +the failed operation for a limited number of times while running +with reduced performance. </p> </dd> + +<dt> Prevention: </dt> <dd> <p> Monitor your LMDB files for +MDB_READERS_FULL errors. After making the necessary adjustments, +restart Postfix. </p> </dd> + +</dl> + +--> + +<!-- + +<p> <strong>Unexpected postmap(1)/postalias(1) "database full" +errors. </strong></p> + +<dl> + +<dt> Problem: </dt> <dd> <p> The "postmap lmdb:filename" command +fails with an MDB_TXN_FULL error. This problem does not exist with +other Postfix databases. </p> </dd> + +<dt> Background: </dt> + +<dd> + +<p> The LMDB implementation has a hard limit on the total transaction +size. This limit is independent of the LMDB database size. Therefore, +the problem cannot be resolved by increasing the lmdb_map_size +value. </p> + +<p> This symptom is indicative of a flawed design. All LMDB data +structures should share the same storage pool so that they can scale +with the database size, and so that all "out of storage" errors are +resolved by increasing the database size. </p> </dd> + +--> + +<!-- + +<p> Problem: </dt> <dd> <p> The "postmap lmdb:filename" command +fails with an MDB_MAP_FULL error. This problem does not exist with +other Postfix databases. </p> </dd> + +<dl> + +<dt> Background: </dt> + +<dd> + +<p> LMDB databases have a hard size limit (configured with the +lmdb_map_size configuration parameter). </p> + +<p> When executing "postmap lmdb:filename", the Postfix LMDB database +client stores the new data in a transaction which takes up space +in addition to the existing data, and commits the transaction when +it closes the database. Only then can the space for old data be +reused. </p> + +</dd> + +<dt> Impact: </dt> <dd> <p> This failure does not affect Postfix +availability, because the old data still exists in the database. +</p> </dd> + +<dt> Mitigation: </dt> <dd> + +<p> When the postmap(1) or postalias(1) command fails with an +MDB_MAP_FULL error, it expands the database file size to the current +LMDB map size limit before terminating. </p> + +<p> Next, when you re-run the postmap(1) or postalias(1) command, +it discovers that the LMDB file is larger than lmdb_map_size/3, +logs a warning, and uses a larger LMDB map size limit instead: </p> + +<p> <tt> warning: <i>filename</i>.lmdb: file size 15024128 ≥ +(lmdb map size limit 16777216)/3<br> warning: <i>filename</i>.lmdb: +using map size limit 45072384</tt> </p> + +<p> By repeating the two steps above you can automate recovery and +avoid the need for human intervention. Just repeat "postmap +lmdb:filename" (up to some limit). After each failure it will use +a 3x larger size limit, and eventually the "database full" error +should disappear. This fails only when the disk is full or when +the LMDB map size limit would exceed the memory address space size +limit. </p> + +<dt> Prevention: </dt> <dd> <p> Monitor your LMDB files and make +sure that in main.cf, lmdb_map_size > 3x the largest LMDB file +size. </p> </dd> </dl> + +</dl> + +--> + +<!-- + +<p> <strong>Unexpected Postfix daemon "database full" errors. +</strong></p> + +<dl> + +<dt> Problem: </dt> <dd> <p> Postfix daemon programs fail with +"database full" errors, such as postscreen(8), tlsmgr(8) or verify(8). +This problem does not exist with other Postfix databases. </p> +</dd> + +<dt> Impact: </dt> <dd> <p> This failure temporarily affects Postfix +availability. The daemon restarts automatically and tries to open +the database again as described next. </p> </dd> + +<dt> Mitigation: </dt> <dd> <p> When a Postfix daemon opens an LMDB +file larger than lmdb_map_size/3, it logs a warning and uses a +larger size limit instead: </p> + +<p> <tt> warning: <i>filename</i>.lmdb: file size 15024128 ≥ +(lmdb map size limit 16777216)/3 <br>warning: <i>filename</i>.lmdb: +using map size limit 45072384</tt> </p> + +<p> This can be used to automate recovery and avoid the need for +human intervention. Each time the daemon runs into a "database full" +error, it restarts and uses a 3x larger size limit. The "database +full" error will disappear, at least for a while. </p> + +<dt> Prevention: </dt> <dd> <p> Monitor your LMDB files and make +sure that lmdb_map_size > 3x the largest LMDB file size. </p> +</dd> </dl> + +--> + +<!-- + +<p> <strong>Non-obvious recovery with postmap(1), postalias(1), or +tlsmgr(8) from a corrupted database. </strong></p> + +<dl> + +<dt> Problem: </dt> <dd> <p> A corrupted LMDB database can't be +rebuilt simply by re-running postmap(1) or postalias(1), or by +waiting until a tlsmgr(8) daemon restarts. This problem does not +exist with other Postfix databases. </p> </dd> + +<dt> Background: </dt> <dd> <p> The Postfix LMDB database client +does not truncate the database file. Instead it attempts to create +a transaction for a "drop" request plus subsequent "store" requests. +That is obviously not possible with a corrupted database file. </p> +</dd> + +<dt> Impact: </dt> <dd> <p> Postfix does not process mail until +someone fixes the problem. </p> </dd> + +<dt> Recovery: </dt> <dd> <p> First delete the ".lmdb" file by hand. +Then rebuild the file with the postmap(1) or postalias(1) +command if the file was created with those commands, or restart +postfix daemons if the file is maintained by tlsmgr(8). +</p> </dd> + +<dt> Prevention: </dt> <dd> + +<p> Arrange your file systems such that they never run out of free +space. </p> + +<p> Use ECC memory to detect and correct silent corruption of +in-memory file system data and metadata. </p> + +<p> Use a file system such as ZFS to detect and correct silent +corruption of on-disk file system data and metadata. DO NOT +use ZFS on systems without ECC memory error correction. </p> + +</dd> </dl> + +--> + +</body> + +</html> |