[389-commits] [389-ds-base] 01/03: Ticket 49267 - autosize split of 0 results in dbcache of 0

This is an automated email from the git hooks/post-receive script.

mreynolds pushed a commit to branch 389-ds-base-1.3.6
in repository 389-ds-base.

commit bd28bdd376c1f467705f7c2e58fc8d62ee0a1e54
Author: William Brown <firstyear@redhat.com>
Date: Tue May 23 11:03:24 2017 +1000

Ticket 49267 - autosize split of 0 results in dbcache of 0

Bug Description: autosize split of 0 results in a dbcache of 0. This was
due to a missing bounds check on the value for 0. In theory this could
still be problematic if the value was say 1% ... But hopefully we don't
see that :)

Fix Description: Add the bounds check.

https://pagure.io/389-ds-base/issue/49267

Author: wibrown

Review by: mreynolds (Thanks!)

(cherry picked from commit 22d4865ea20acb6e6c11aed10d09241b09bb711c)
---
ldap/servers/slapd/back-ldbm/start.c | 14 +++++++++++++-
1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/ldap/servers/slapd/back-ldbm/start.c b/ldap/servers/slapd/back-ldbm/start.c
index a207bd8..1834a19 100644
--- a/ldap/servers/slapd/back-ldbm/start.c
+++ b/ldap/servers/slapd/back-ldbm/start.c
@@ -101,7 +101,11 @@ ldbm_back_start_autotune(struct ldbminfo *li) {
/* This doesn't control the availability of the feature, so we can take the
* default from ldbm_config.c
*/
- autosize_db_percentage_split = li->li_cache_autosize_split;
+ if (li->li_cache_autosize_split == 0) {
+ autosize_db_percentage_split = 40;
+ } else {
+ autosize_db_percentage_split = li->li_cache_autosize_split;
+ }


/* Check the values are sane. */
@@ -131,10 +135,18 @@ ldbm_back_start_autotune(struct ldbminfo *li) {
db_size = (autosize_db_percentage_split * zone_size) / 100;

/* Cap the DB size at 512MB, as this doesn't help perf much more (lkrispen's advice) */
+ /* NOTE: Do we need a minimum DB size? */
if (db_size > (512 * MEGABYTE)) {
db_size = (512 * MEGABYTE);
}

+ /* NOTE: Because of how we workout entry_size, even if
+ * have autosize split to say ... 90% for dbcache, because
+ * we cap db_size, we use zone_size - db_size, meaning that entry
+ * cache still gets the remaining memory *even* though we didn't use it all.
+ * If we didn't do this, entry_cache would only get 10% of of the avail, even
+ * if db_size was caped at say 5% down from 90.
+ */
if (backend_count > 0 ) {
/* Number of entry cache pages per backend. */
entry_size = (zone_size - db_size) / backend_count;

--
To stop receiving notification emails like this one, please contact
the administrator of this repository.
_______________________________________________
389-commits mailing list -- 389-commits@lists.fedoraproject.org
To unsubscribe send an email to 389-commits-leave@lists.fedoraproject.org