From: Nikita Pettik <korablev@tarantool.org> To: tarantool-patches@dev.tarantool.org Cc: v.shpilevoy@tarantool.org Subject: [Tarantool-patches] [PATCH] region: do not rotate slabs in case of single slab purification Date: Thu, 6 Feb 2020 20:14:37 +0300 [thread overview] Message-ID: <d4db4ccd7c65911eb17b7d6816ed8f8c663b5940.1581008805.git.korablev@tarantool.org> (raw) Imagine following situation at the moment of region truncation: 1 slab 2 slab/HEAD x of s1 used y of s2 used +---------+ +---------+ ...->|xx| |-->|yyyyyy| |--> NULL |xx| | |yyyyyy| | +---------+ +---------+ And region is going to be truncated by size y (i.g. region_truncate(®ion, region_used() - y)). After this operation second slab will remain empty (slab.used == 0). Currently, this slab is returned back to the slab cache, which leads to slab rotation. This may not be useful in terms of performance, especially for requests which don't abuse region memory (for instance, SQL queries) but do call region_truncate() intensively. For example, during transfer tuples to ephemeral table truncate() is called for each tuple. What is more, truncate is also called on each returned tuple in sql_row_to_port() (see runtime_tuple_new()). Also, taking into consideration revert of 67d7ab4 (see 81dda5b) this may lead to memory usage fluctuations: region memory in use before query execution and after may not be the same. For instance: total region memory in usage is x, but the first slab (head) is empty (since between requests region_reset() is called). During query execution region_truncate() is called and first slab is put back to the slab cache. Now another slab forms head of slab list, but there's no guarantee that its size is 0 (since slab's sizes are not zeroed on reset). As a result, total region memory may change (decrease) due to slab rotation. It is likely to be misleading for users and in particular for our testing system: to detect region memory leak cases, we have nothing to do but restart Tarantool instance each time before test. --- Branch: https://github.com/tarantool/small/tree/np/dont-rotate-region-on-truncate small/region.c | 2 +- test/region.c | 69 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 70 insertions(+), 1 deletion(-) diff --git a/small/region.c b/small/region.c index 04476c4..b60ad38 100644 --- a/small/region.c +++ b/small/region.c @@ -79,7 +79,7 @@ region_truncate(struct region *region, size_t used) struct rslab *slab = rlist_first_entry(®ion->slabs.slabs, struct rslab, slab.next_in_list); - if (slab->used > cut_size) { + if (slab->used >= cut_size) { /* This is the last slab to trim. */ slab->used -= cut_size; cut_size = 0; diff --git a/test/region.c b/test/region.c index bd5a4c1..8735a59 100644 --- a/test/region.c +++ b/test/region.c @@ -63,6 +63,74 @@ region_test_truncate() footer(); } +/** + * Make sure that in case cut_size == slab_used (i.e. we are going + * to truncate whole slab) slab itself is not put back to the slab + * cache. It should be kept in the list and reused instead. + */ +void +region_test_truncate_rotate() +{ + header(); + + struct region region; + + region_create(®ion, &cache); + + size_t alloc1_sz = 10; + void *ptr = region_alloc(®ion, alloc1_sz); + + fail_unless(ptr != NULL); + + fail_if(region_total(®ion) > 5000); + size_t total_before_alloc = region_total(®ion); + + /* + * This allocation does not fit in previously + * allocated slab (default slab size is 4kb), so + * there's two slabs now in the region list. + */ + size_t alloc2_sz = 10000; + region_alloc(®ion, alloc2_sz); + fail_if(total_before_alloc == region_total(®ion)); + + /* + * Before truncate ('x' is occupied space): + * + * 1 slab 2 slab/HEAD + * 10b used 10kb used + * +---------+ +---------+ + * |xx| |-->|xxxxxx| |--> NULL + * |xx| | |xxxxxx| | + * +---------+ +---------+ + * + * After truncate: + * + * 1 slab 2 slab/HEAD + * 10b used 0kb used + * +---------+ +---------+ + * |xx| |-->| |--> NULL + * |xx| | | | + * +---------+ +---------+ + * + */ + size_t total_before_truncate = region_total(®ion); + region_truncate(®ion, alloc1_sz); + + /* + * Whole slab has been purified but it shouldn't be + * returned back to the slab cache. So that it can be + * reused for further allocations. + */ + fail_if(total_before_truncate != region_total(®ion)); + region_alloc(®ion, alloc2_sz); + fail_if(total_before_truncate != region_total(®ion)); + + region_free(®ion); + + footer(); +} + int main() { quota_init("a, UINT_MAX); @@ -72,6 +140,7 @@ int main() region_basic(); region_test_truncate(); + region_test_truncate_rotate(); slab_cache_destroy(&cache); } -- 2.15.1
next reply other threads:[~2020-02-06 17:14 UTC|newest] Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top 2020-02-06 17:14 Nikita Pettik [this message] 2020-02-06 18:43 ` Konstantin Osipov 2020-02-06 20:48 ` Vladislav Shpilevoy 2020-02-07 13:58 ` Nikita Pettik 2020-02-08 0:28 ` Vladislav Shpilevoy 2020-02-10 11:52 ` Nikita Pettik 2020-04-10 10:10 Aleksandr Lyapunov
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=d4db4ccd7c65911eb17b7d6816ed8f8c663b5940.1581008805.git.korablev@tarantool.org \ --to=korablev@tarantool.org \ --cc=tarantool-patches@dev.tarantool.org \ --cc=v.shpilevoy@tarantool.org \ --subject='Re: [Tarantool-patches] [PATCH] region: do not rotate slabs in case of single slab purification' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox