On Mon, Mar 25, 2019 at 4:57 PM Ben Pfaff <[email protected]> wrote:
>
> I was seeing a lot of these messages, including a lot of them suppressed
> by rate-limiting, and I wondered whether any really big messages were
> being suppressed.  By reporting the largest bucket, instead of just every
> large bucket, it becomes more likely that the truly too-large buckets get
> reported.
>
> (The problem I saw was a false alarm.)
>
> Signed-off-by: Ben Pfaff <[email protected]>
> ---
>  lib/hmap.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/lib/hmap.c b/lib/hmap.c
> index 1ba4a5716a78..20161698af5d 100644
> --- a/lib/hmap.c
> +++ b/lib/hmap.c
> @@ -1,5 +1,5 @@
>  /*
> - * Copyright (c) 2008, 2009, 2010, 2012, 2013, 2015 Nicira, Inc.
> + * Copyright (c) 2008, 2009, 2010, 2012, 2013, 2015, 2019 Nicira, Inc.
>   *
>   * Licensed under the Apache License, Version 2.0 (the "License");
>   * you may not use this file except in compliance with the License.
> @@ -103,6 +103,7 @@ resize(struct hmap *hmap, size_t new_mask, const char 
> *where)
>              tmp.buckets[i] = NULL;
>          }
>      }
> +    int max_count = 0;
>      for (i = 0; i <= hmap->mask; i++) {
>          struct hmap_node *node, *next;
>          int count = 0;
> @@ -111,15 +112,20 @@ resize(struct hmap *hmap, size_t new_mask, const char 
> *where)
>              hmap_insert_fast(&tmp, node, node->hash);
>              count++;
>          }
> -        if (count > 5) {
> -            static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10);
> -            COVERAGE_INC(hmap_pathological);
> -            VLOG_DBG_RL(&rl, "%s: %d nodes in bucket (%"PRIuSIZE" nodes, 
> %"PRIuSIZE" buckets)",
> -                        where, count, hmap->n, hmap->mask + 1);
> +        if (count > max_count) {
> +            max_count = count;
>          }
>      }
>      hmap_swap(hmap, &tmp);
>      hmap_destroy(&tmp);
> +
> +    if (max_count > 5) {
> +        static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(10, 10);
> +        COVERAGE_INC(hmap_pathological);
> +        VLOG_DBG_RL(&rl, "%s: %d nodes in bucket "
> +                    "(%"PRIuSIZE" nodes, %"PRIuSIZE" buckets)",
> +                    where, max_count, hmap->n, hmap->mask + 1);
> +    }
>  }

Would it be more helpful in the same log print out how many buckets
has count > 5? For example, 1 entry with count = 10 maybe is not a big
deal, but too many entries with count > 5 may indicate bad hashing.
The new log may not tell this information.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to