Re: Too long startup time after each crash.

On 12/21/2016 02:17 PM, neos(at)olansoft(dot)com wrote:
> Hi.
>
> Perhaps i've sent it to wrong mail-list previously.
>
> After the upgrade postgresql had a strange behavior. After the crash, server launch now take a long time, up to 30-40 minutes sometimes. When connecting to the server (as well as in the log), the message - "the database system is starting up". Startup process consumes 0.4 - 1% CPU.

What upgrade and how was it done?

What crash and what caused it?

In particular why are you seeing multiple crashes?

What is the log just before the crash?

What is the log after the the "database system is starting up" part?

>
> Tried different version 9.5.[2,3,4], 9.6.1 (9.4 works fine). On the Internet this problem is not met, perhaps the problem lies in a combination of different factors (configuration database, Linux kernel), may be my config is wrong?

Do you mean you searched for this problem description and could not find it?

>
> Technical details:
>
> Gentoo Linux 4.4.2-hardened
> current pg version 9.6.1
>
> ## pg config:
>
> listen_addresses = '*'
> port = 5432
> max_connections = 500
> shared_buffers = 2048MB
> temp_buffers = 16MB
> work_mem = 64MB
> maintenance_work_mem = 512MB
> max_stack_depth = 4MB
> dynamic_shared_memory_type = posix
> wal_level = replica
> wal_buffers = 8MB
> checkpoint_completion_target = 0.85
> archive_mode = on
> archive_command = 'gzip < %p > /home/pgbasebackup/xlogs/%f.gz'
> max_wal_senders = 3
> effective_cache_size = 9GB
> log_min_duration_statement = 1500
> log_timezone = 'Asia/Yekaterinburg'
> datestyle = 'iso, dmy'
> timezone = 'Asia/Yekaterinburg'
> lc_messages = 'ru_RU.UTF-8'
> lc_monetary = 'ru_RU.UTF-8'
> lc_numeric = 'ru_RU.UTF-8'
> lc_time = 'ru_RU.UTF-8'
> default_text_search_config = 'pg_catalog.russian'
> plperl.on_init = 'use utf8; use re; package utf8; require "utf8_heavy.pl";'
>
> ## end of config
>
> Sorry for my eng :)
>

--
Adrian Klaver
adrian(dot)klaver(at)aklaver(dot)com