syncrepl consumer is slow
by Howard Chu
One thing I just noticed, while testing replication with 3 servers on my
laptop - during a refresh, the provider gets blocked waiting to write to
the consumers after writing about 4000 entries. I.e., the consumers
aren't processing fast enough to keep up with the search running on the
provider.
(That's actually not too surprising since reads are usually faster than
writes anyway.)
The consumer code has lots of problems as it is, just adding this note
to the pile.
I'm considering adding an option to the consumer to write its entries
with dbnosync during the refresh phase. The rationale being, there's
nothing to lose anyway if the refresh is interrupted. I.e., the consumer
can't update its contextCSN until the very end of the refresh, so any
partial refresh that gets interrupted is wasted effort - the consumer
will always have to start over from the beginning on its next refresh
attempt. As such, there's no point in safely/synchronously writing any
of the received entries - they're useless until the final contextCSN update.
The implementation approach would be to define a new control e.g. "fast
write" for the consumer to pass to the underlying backend on any write
op. We would also have to e.g. add an MDB_TXN_NOSYNC flag to
mdb_txn_begin() (BDB already has the equivalent flag).
This would only be used for writes that are part of a refresh phase. In
persist mode the provider and consumers' write speeds should be more
closely matched so it wouldn't be necessary or useful.
Comments?
--
-- Howard Chu
CTO, Symas Corp. http://www.symas.com
Director, Highland Sun http://highlandsun.com/hyc/
Chief Architect, OpenLDAP http://www.openldap.org/project/
8 years
ITS8100: What to do about a fresh accesslog DB when in delta-sync MMR node
by Quanah Gibson-Mount
There's a general issue with using delta-sync MMR that needs some
resolution (It's causing some considerable pain for customers): It is
possible to put the entire system into endless fallbacks until a new and/or
reloaded node gets a write operation. The general problem is that when an
existing master queries the new/reloaded master's accesslog DB, zero
entries are returned. This then triggers the fallback. This happens up
util such a time as the new/reloaded master gets a direct write op. I've
worked around it in general by immediately doing a no-op on the primary db
(ldapmodify/replace an attribute value with its own value), but it would be
nice to be able to bring new MMR nodes online or be able to reload MMR
nodes for if they get out of sync, etc, without causing this sync fallback
issue. There is no clear solution at the moment on what to do about zero
results from the accesslog in the delta-sync MMR scenario. Proposals
welcome. ;)
--Quanah
--
Quanah Gibson-Mount
Platform Architect
Zimbra, Inc.
--------------------
Zimbra :: the leader in open source messaging and collaboration
8 years, 1 month
RE24 testing call #4 (2.4.41), LMDB RE0.9 testing call #4 (0.9.15)
by Quanah Gibson-Mount
OpenLDAP 2.4.41 Engineering
Fixed ldapsearch to explicitly flush its buffer (ITS#8118)
Fixed libldap async connections (ITS#8090)
Fixed libldap double free of request during abandon (ITS#7967)
Fixed libldap error string for LDAP_X_CONNECTING (ITS#8093)
Fixed libldap segfault in ldap_sync_initialize (ITS#8001)
Fixed libldap ldif-wrap off by one error (ITS#8003)
Fixed libldap handling of TLS in async mode (ITS#8022)
Fixed libldap null pointer dereference (ITS#8028)
Fixed libldap mutex handling with LDAP_OPT_SESSION_REFCNT (ITS#8050)
Fixed slapd slapadd onetime leak with -w (ITS#8014)
Fixed slapd sasl auxprop crash with invalid config (ITS#8092)
Fixed slapd syncrepl delta-mmr issue with overlays and slapd.conf
(ITS#7976)
Fixed slapd syncrepl mutex for cookie state (ITS#7968)
Fixed slapd syncrepl memory leaks (ITS#8035)
Fixed slapd syncrepl to free presentlist at end of refresh mode
(ITS#8038)
Fixed slapd syncrepl to streamline presentlist (ITS#8042)
Fixed slapd syncrepl concurrency when CHECK_CSN is enabled
(ITS#8120)
Fixed slapd rootdn checks for hidden backends (ITS#8108)
Fixed slapd segfault when using matched values control (ITS#8046)
Fixed slapd-mdb minor case typo (ITS#8049)
Fixed slapd-mdb one-level search (ITS#7975)
Fixed slapd-mdb heap corruption (ITS#7965)
Fixed slapd-mdb crash after deleting in-use schema (ITS#7995)
Fixed slapd-mdb minor code cleanup (ITS#8011)
Fixed slapd-mdb to return errors when using incorrect env flags
(ITS#8016)
Fixed slapd-mdb to correctly update search candidates (ITS#8036,
ITS#7904)
Fixed slapd-mdb when there were more than 65535 aliases in scope
(ITS#8103)
Fixed slapd-meta TLS initialization with ldaps URIs (ITS#8022)
Fixed slapd-perl conversion to cn=config (ITS#8105)
Fixed slapo-collect segfault (ITS#7797)
Fixed slapo-constraint with 0 count constraint (ITS#7780,ITS#7781)
Fixed slapo-deref with empty attribute list (ITS#8027)
Fixed slapo-memberof to correctly reject invalid members (ITS#8107)
Fixed slapo-sock result parser for CONTINUE (ITS#8048)
Fixed slapo-syncprov synprov_matchops usage of test_filter
(ITS#8013)
Fixed slapo-syncprov segfault on disconnect/abandon
(ITS#5452,ITS#8012)
Fixed slapo-syncprov memory leak (ITS#8039)
Fixed slapo-syncprov segfault on disconnect/abandon (ITS#8043)
Fixed slapo-syncprov deadlock when autogroup is in use (ITS#8063)
Fixed slapo-syncprov potential loss of changes when under load
(ITS#8081)
Fixed slapo-unique enforcement of uniqueness with manageDSAit
control (ITS#8057)
Build Environment
Fixed libdb detection with gcc 5.x (ITS#8056)
Enhanced contrib modules build paths (ITS#7782)
Fixed contrib/autogroup internal operation identity
(ITS#8006)
Fixed contrib/autogroup to skip internal ops with accesslog
(ITS#8065)
Fixed contrib/passwd/sha2 compiler warning (ITS#8000)
Fixed contrib/noopsrch compiler warning (ITS#7998)
Fixed contrib/dupent compiler warnings (ITS#7997)
Test suite: Added vrFilter test (ITS#8046)
Contrib
Added pbkdf2 sha256 and sha512 schemes (ITS#7977)
Fixed autogroup modification callback responses (ITS#6970)
Fixed nssov compare with usergroup (ITS#8079)
Fixed nssov password change behavior (ITS#8080)
Fixed nssov updated to 0.9.4 (ITS#8097)
Documentation
Added ldap_get_option(3) LDAP_FEATURE_INFO_VERSION
information (ITS#8032)
Added ldap_get_option(3) LDAP_OPT_API_INFO_VERSION
information (ITS#8032)
Fixed slapd-config(5), slapd.conf(5) tls_cipher_suite
option (ITS#8099)
Fixed slapd-meta(5), slapd-ldap(5) tls_cipher_suite option
(ITS#8099)
LMDB 0.9.15 Release Engineering
Fix txn init (ITS#7961,#7987)
Fix MDB_PREV_DUP (ITS#7955,#7671)
Fix compact of empty env (ITS#7956)
Fix mdb_rebalance collapsing root (ITS#8062)
Fix mdb_load with large values (ITS#8066)
Fix to retry writes on EINTR (ITS#8106)
Fix mdb_cursor_del on empty DB (ITS#8109)
Fix MDB_INTEGERDUP key compare (ITS#8117)
Added workaround for fdatasync bug in ext3fs
Build
Don't use -fPIC for static lib
Update .gitignore (ITS#7952,#7953)
Cleanup for "make test" (ITS#7841)
Misc. Android/Windows cleanup
Documentation
Fix MDB_APPEND doc
Clarify mdb_dbi_open doc
Thanks!
--Quanah
--
Quanah Gibson-Mount
Platform Architect
Zimbra, Inc.
--------------------
Zimbra :: the leader in open source messaging and collaboration
8 years, 1 month
MDB_INTEGERDUP in back-mdb's indexer for attr-tables.
by Леонид Юрьев
Hi, Howard.
I would like to refine my question.
Currently I develop a mdb_chk tool for deep check LMDB-files, like a fsck tool.
It is needed for us, but let this stay aside for now.
On some step I plan to check the size of each key in each DBI with
MDB_INTEGERKEY and/or MDB_INTEGERDUP flags.
Initially I assumed - if one of these flags is set, the the length of
all keys must be sizeof(int) or sizeof(long). As will be used the
corresponding comparison functions (mdb_cmp_int, mdb_cmp_long) , which
can handle only keys of such sizes.
But this approach has failed. A dbi-tables created for attr by
mdb_attr_dbs_open() have MDB_DUPSORT|MDB_DUPFIXED|MDB_INTEGERDUP flags
and may contain records with a 5-byte keys. For instance as the
'entryCSN'.
Later you have written that MDB_INTEGERDUP is required.
Now I do not understand how it all works:
- MDB_INTEGERDUP is set for attr tables.
- indexer could create a 5-byte keys.
- mdb-backend set custom comparison function only for MDB_DN2ID table.
- under debugger I see that md_dcmp is pointed to the default.
- assert for key-size in default comparators was not failed while "make test".
- all test are passed without MDB_INTEGERDUP.
- but you told MDB_INTEGERDUP is necessary
So,
- how and when MDB_INTEGERDUP will have an impact?
- in which cases keys in the table definitely should be sizeof(int) or
sizeof(long)?
Thank in advance,
Leonid.
8 years, 1 month
Re: (ITS#8117) Bugs related to key-size in lmdb and backend
by Леонид Юрьев
Hi, Howard.
As you told before, I will write to openldap-devel.
So, please take look to MDB_INTEGERDUP in mdb_attr_dbs_open().
I cutout ones and all tests are passing.
For my mdb_chk tool this is also solves a problem with checking
INTEGER-keys.
With MDB_INTEGERDUP removal a logic become spotless: keys must be
sizeof(int) or sizeof(size_t) only if MDB_INTEGERDUP or MDB_INTEGERKEY is
in a dbi-flags.
Would you agree to remove MDB_INTEGERDUP from the mdb_attr_dbs_open()?
Leonid.
2015-04-30 4:12 GMT+03:00 Howard Chu <hyc(a)symas.com>:
> Leonid Yuriev wrote:
>
>> OK, but then - what is the purpose of MDB_INTEGERDUP in
>> mdb_attr_dbs_open()?
>>
>> flags = MDB_DUPSORT|MDB_DUPFIXED|MDB_INTEGERDUP;
>> if ( !(slapMode & SLAP_TOOL_READONLY) )
>> flags |= MDB_CREATE;
>>
>> Think it is not necessary, is it?
>>
>
> You really think so? Have you thought about what the actual *values* are
> in an index, instead of just the keys?
>
> Anyway, this ITS is resolved. Thank you for reporting the bug. If you want
> to ask further questions about how the indexer works take it to
> openldap-devel.
>
>
> --
> -- Howard Chu
> CTO, Symas Corp. http://www.symas.com
> Director, Highland Sun http://highlandsun.com/hyc/
> Chief Architect, OpenLDAP http://www.openldap.org/project/
>
8 years, 1 month
