Skip to contentSkip to conversation

cooker: use BB_HASHSERVE_DB_DIR for hash server database location

PATCH Superseded Alexander Kanavin 1 participant 2 messages Jan 29

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 cooker: use BB_HASHSERVE_DB_DIR for hash server database location

From: Alexander Kanavin <[email protected]>

If unset, the existing behavior is preseved.

The use case is sharing the database in bitbake-setup's driven builds without having to set up/start/stop a common single server shared between them (this is added to bitbake-setup in the next commit).

Also create the specified directory if it doesn't yet exist.

Add a check that the directory is not on a NFS path, and error out then, guiding the user to set up a standalone hash equivalency server.

Note: the check runs 'stat' executable with a format string parameter, ensuring it prints only the filesystem type and nothing else. Python's standard library does not have a statfs() wrapper, and using ctypes to call into it would've required delicate, crash-prone data type definitions, full of magic numbers.

---
 .../bitbake-user-manual-ref-variables.rst     | 23 +++++++++++++++++++
 lib/bb/cooker.py                              | 11 ++++++++-
 lib/bb/utils.py                               |  8 +++++++
 3 files changed, 41 insertions(+), 1 deletion(-)
doc/bitbake-user-manual/bitbake-user-manual-ref-variables.rst +23 -0
index af911f306..4b3b10d46 100644
@@ -475,6 +475,29 @@ overview of their function and contents.
       equivalences that correspond to Share State caches that are
       only available on specific clients.
 
+   :term:`BB_HASHSERVE_DB_DIR`
+      When :term:`BB_HASHSERVE` is set to ``auto``, bitbake will use
+      a private location inside a particular build directory for the sqlite
+      database file that holds hash equivalency data.
+
+      This variable allows using a different path, which can be shared
+      between multiple build directories and bitbake instances
+      that operate on them. This enables using a common ``${SSTATE_DIR}``
+      together with common hash equivalency data for local builds, without having to
+      separately manage a hash equivalency server.
+
+      .. note::
+
+         This variable cannot be set to a NFS mount and bitbake will error out then.
+         The reason is that NFS implementations can have file locking issues, which
+         can cause data loss and corruption when there are multiple writers operating
+         on a file at the same time as explained in https://sqlite.org/faq.html#q5
+
+         If you'd like to share hash equivalency data between multiple computers, you
+         need to set up a hash equivalency server separately and point :term:`BB_HASHSERVE`
+         to it. See https://docs.yoctoproject.org/dev-manual/hashequivserver.html for
+         additional information.
+
    :term:`BB_HASHSERVE_UPSTREAM`
       Specifies an upstream Hash Equivalence server.
 
lib/bb/cooker.py +10 -1
index fa7747d64..4b6ba3196 100644
@@ -328,7 +328,16 @@ class BBCooker:
         if self.data.getVar("BB_HASHSERVE") == "auto":
             # Create a new hash server bound to a unix domain socket
             if not self.hashserv:
-                dbfile = (self.data.getVar("PERSISTENT_DIR") or self.data.getVar("CACHE")) + "/hashserv.db"
+                bb_hashserve_db_dir = self.data.getVar("BB_HASHSERVE_DB_DIR")
+                if bb_hashserve_db_dir and utils.is_path_on_nfs(bb_hashserve_db_dir):
+                    bb.fatal("""Hash equivalency database location (set via BB_HASHSERVE_DB_DIR to {})
+cannot be on a NFS mount due to potential NFS locking issues between sqlite clients, per https://sqlite.org/faq.html#q5
+
+If you need to share the database between several computers, set up a permanently running hash equivalency server
+according to https://docs.yoctoproject.org/dev-manual/hashequivserver.html""".format(bb_hashserve_db_dir))
+                dbdir = bb_hashserve_db_dir or self.data.getVar("PERSISTENT_DIR") or self.data.getVar("CACHE")
+                os.makedirs(dbdir, exist_ok=True)
+                dbfile = dbdir + "/hashserv.db"
                 upstream = self.data.getVar("BB_HASHSERVE_UPSTREAM") or None
                 if upstream:
                     try:
lib/bb/utils.py +8 -0
index 366836bfc..974611bc7 100644
@@ -2264,3 +2264,11 @@ def lock_timeout_nocheck(lock):
         if l:
             lock.release()
         sigmask(signal.SIG_SETMASK, s)
+
+def is_path_on_nfs(path):
+    """
+    Returns True if ``path`` argument is on a NFS mount.
+    """
+    import bb.process
+    fstype = bb.process.run("stat -f -c %T {}".format(path))[0].strip()
+    return fstype == "nfs"
Signature · 1 line
2.47.3
#1 lore raw

PATCH v2 2/2 bitbake-setup: share sstate by default between builds

From: Alexander Kanavin <[email protected]>

Nowadays sharing sstate must also include sharing the hash equivalency information and thus, managing a hash equivalency server. There are two ways to do it:

- starting/stopping the server outside the bitbake invocations, and guaranteeing that it's available when bitbake is invoked.

- using bitbake's built-in start/stop code which launches a server before a build starts and stops it when a build is finished; essentially this is a private server, using a database private to a build directory (by default).

I couldn't come up with a good way to do the first option in bitbake-setup:
it needs to be invisible to users, they should not have to run special commands and they should not wonder why there is a mysterious background process. It's not impossible to auto-start a shared server, but that will quickly run into synchronization issues: if one server is being started, another should not be started at the same time. If one server is shutting down (e.g. after an inactivity timeout), another starting server should wait until it frees the socket, and block all bitbake invocations on that. Memory resident bitbake does this in lib/bb/server/process.py with a lot of complexity, and I don't think it should be added to the hash server as well.

On the other hand, hash equivalency database is sqlite-driven, and sqlite documentation reassures that sharing it between different simultaneous processes is okay: nothing will get lost or corrupted: https://sqlite.org/faq.html#q5

I've confirmed this by running simultaneous builds that way: nothing unusual happened, and sstate was shared as it's supposed to.

There's a new setting that turns off this behavior for situations where the server and sstate are managed externally.

---
v2: use ?= (weak assignments) for sstate variables to make it easier to override them in particular builds, set hashserve db location from value of SSTATE_DIR, rather than as direct value.

---
 bin/bitbake-setup                             | 28 ++++++++++++++++++-
 .../bitbake-user-manual-environment-setup.rst | 19 +++++++++++++
 2 files changed, 46 insertions(+), 1 deletion(-)
bin/bitbake-setup +27 -1
index 7f6ea550c..679654336 100755
@@ -838,6 +838,31 @@ def create_siteconf(top_dir, non_interactive, settings):
 
         os.makedirs(top_dir, exist_ok=True)
         with open(siteconfpath, 'w') as siteconffile:
+            sstate_settings = textwrap.dedent(
+                """
+                #
+                # Where to place shared-state files
+                #
+                # BitBake has the capability to accelerate builds based on previously built output.
+                # This is done using "shared state" files which can be thought of as cache objects
+                # and this option determines where those files are placed.
+                #
+                # You can wipe out TMPDIR leaving this directory intact and the build would regenerate
+                # from these files if no changes were made to the configuration. If changes were made
+                # to the configuration, only shared state files where the state was still valid would
+                # be used (done using checksums).
+                SSTATE_DIR ?= "{sstate_dir}"
+                #
+                # Hash Equivalence database location
+                #
+                # Hash equivalence improves reuse of sstate by detecting when a given sstate
+                # artifact can be reused as equivalent, even if the current task hash doesn't
+                # match the one that generated the artifact. This variable controls where the
+                # Hash Equivalence database ("hashserv.db") is stored and can be shared between
+                # concurrent builds.
+                BB_HASHSERVE_DB_DIR ?= "${SSTATE_DIR}"
+                """.format(sstate_dir=os.path.join(top_dir, ".sstate-cache"))
+            )
             siteconffile.write(
                 textwrap.dedent(
                     """\
@@ -855,7 +880,7 @@ def create_siteconf(top_dir, non_interactive, settings):
                     """.format(
                         dl_dir=settings["default"]["dl-dir"],
                     )
-                )
+                ) + (sstate_settings if settings["default"]["common-sstate"] == 'yes' else "")
             )
 
 
@@ -1061,6 +1086,7 @@ def main():
                          'top-dir-name':'bitbake-builds',
                          'registry':default_registry,
                          'use-full-setup-dir-name':'no',
+                         'common-sstate':'yes',
                          }
 
         global_settings = load_settings(global_settings_path(args))
doc/bitbake-user-manual/bitbake-user-manual-environment-setup.rst +19 -0
index 3b6a73fd8..b7212f1af 100644
@@ -517,6 +517,7 @@ A valid settings file would for example be:
    registry = /path/to/bitbake/default-registry
    dl-dir = /path/to/bitbake-setup-downloads
    use-full-setup-dir-name = yes
+   common-sstate = yes
 
 Settings and their values can be listed and modified with the ``bitbake-setup
 settings`` command. See the :ref:`ref-bbsetup-command-settings` section for
@@ -621,6 +622,24 @@ will override the suggestions for the :term:`Setup` directory name made by
 will make the directory names longer, but fully specific: they will contain
 all selections made during initialization.
 
+.. _ref-bbsetup-setting-common-sstate:
+
+``common-sstate``
+-----------------
+
+When this setting is set to ``yes`` (which is also the default), bitbake-setup will
+set up a common sstate directory and common hash equivalency database for all the
+:term:`setups <Setup>` in a :term:`Top Directory`. This is very beneficial for speeding
+up builds as build artefacts will be reused whenever possible between them.
+
+Set this to ``no`` for advanced use cases, such as placing the sstate directory on a NFS
+mount and maintaining a separate hash equivalency server, so that sstate and hash equivalency
+data can be shared between several computers. For such use cases the sstate settings need
+to be added to a build configuration separately.
+
+See https://docs.yoctoproject.org/dev-manual/hashequivserver.html for how to share sstate
+on the network.
+
 .. _ref-bbsetup-section-config-reference:
 
 Generic Configuration Files Reference
Signature · 1 line
2.47.3