vagetablechicken

tablet配置变更或升级，统一都是用pre-upgrade和post-upgrade。
**迁移过程中需要停写，否则可能有数据丢失。**
注意集群中的table特点，分两种情况：
1. 有大量table是单副本，pre-upgrade是要把分片挪给别的机器，中途内存占用会变多，可能迁的途中导致OOM。内存吃紧时，
可以新建tablet server，pre-upgrade每次都是挑replica总数最少的机器，基本上就是p2p地传到新tablet server上。不用担心其他tablet会上涨很多。
2. 大部分table都是多副本表，无需担心内存，本质上只做changeleader，不涉及数据迁移。但注意，不停写的话，数据可能丢失。

步骤：
1. 通过showtablestatus来确认集群现状，参考 https://openmldb.ai/docs/zh/main/maintain/openmldb_ops.html。注意请使用0.9.0以上的tools，对0.8.x兼容。重点看Replica列，是1还是大于1。
2. 确认是上文的哪种情况，大部分表是单副本还是多副本。

vagetablechicken

stdTTL: to avoid ttl like absorlat(10,0), the simpler form should be abs(10). If not standard, we may get too large merged ttl, otherwise we should check the values if == 0. Standard is better.

so after std, we may get two ttl, abs(0), abs(x), lat(0), lat(x), absandlat(x,y), absorlat(x,y), x>0 and y>0.

lat(0) to abs(0)?

If two ttl are the same type, it's ok to merge the value, 0 means never exipres.

If different types, I'll try my best to set a small ttl to cover the two ttl.

vagetablechicken

{
  "__inputs": [
    {
      "name": "DS_PROMETHEUS",
      "label": "prometheus",
      "description": "",
      "type": "datasource",
      "pluginId": "prometheus",
      "pluginName": "Prometheus"
    }

vagetablechicken

#include <dlfcn.h>
#include <stdio.h>

int main(void) {
    void *handle;
    char *error;

    handle = dlopen("libtest_udf.so", RTLD_LAZY);
    if (!handle) {
        fprintf(stderr, "%s\n", dlerror());

vagetablechicken

create libopenmldbsdk.a 
addlib /home/huangwei/OpenMLDB/build/src/sdk/libopenmldb_api.a 
addlib /home/huangwei/OpenMLDB/build/src/libopenmldb_sdk.a 
addlib /home/huangwei/OpenMLDB/build/src/libopenmldb_codec.a 
addlib /home/huangwei/OpenMLDB/build/src/libopenmldb_catalog.a 
addlib /home/huangwei/OpenMLDB/build/src/libschema.a 
addlib /home/huangwei/OpenMLDB/build/src/libclient.a 
addlib /home/huangwei/OpenMLDB/build/src/libzk_client.a 
addlib /home/huangwei/OpenMLDB/build/src/libbase.a 
addlib /home/huangwei/OpenMLDB/build/src/libopenmldb_flags.a 

vagetablechicken

import requests
import json
# detect table status about data

db = 'disk_test'
ns_leader = '172.24.4.27:7123'

url = f'http://{ns_leader}/NameServer/ShowTable'
res = requests.get(url, json={"show_all": True})
tables = json.loads(res.text)['table_info']

vagetablechicken

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
# Copyright 2021 4Paradigm
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#

vagetablechicken

openmldb/bin/openmldb --zk_cluster=127.0.0.1:2181 --zk_root_path=/openmldb --role=sql_client < test.sql 即可 注意test.sql内，c1.csv的路径，需要跟taskmanager（spark.home=local）在同一台机器上，并且写绝对路径。 输出结果里都是SUCCEED，没有ERROR就可以。

vagetablechicken

higher gcc(>=7) don't have pthread,m,...
gcc 4.8.5/5.4.0 download and their prerequire dependencies:
https://gcc.gnu.org/pub/gcc/
depends: https://gcc.gnu.org/pub/gcc/infrastructure/

mkdir build & cd build
../configure --prefix=`pwd`/../output --enable-checking=release --enable-languages=c,c++ --disable-multilib # Multilib means support for multiple architectures

make
make install

vagetablechicken

import math


class WinnerTree:
    def __init__(self, ext):
        self.ext = ext
        self.N = len(ext)
        self.k = math.ceil(math.log(self.N, 2))
        print(self.k, ", bottom layer node ", pow(2, self.k))