etcd基本使用

CRUD

1、基本的put/get/del

#普通的get与delete
res = etcd.put('bar')
value, kv_msg = etcd.get('bar')
print(value)
print(etcd.delete('bar'))
print(etcd.get('bar'))

2、获取当前所有的key

print(etcd.put('demo/bar1', 'doot'))
print(etcd.put('demo/bar2', 'doot'))
list_of_kv = list(etcd.get_all())
for value, kv_msg in list_of_kv:
    print("key:{}, value:{}" .format(kv_msg.key, value))

3、获取/删除带有前缀的键

# [2]获取带有前缀的键
list_of_kv = list(etcd.get_prefix('demo/'))
for value, kv_msg in list_of_kv:
    print("key:{}, value:{}" .format(kv_msg.key, value))
# [3]删除带有前缀的键
res = etcd.delete_prefix('demo/')
print(res)
list_of_kv = list(etcd.get_prefix('demo/'))
for value, kv_msg in list_of_kv:
    print("key:{}, value:{}" .format(kv_msg.key, value))

lease使用

1、创建lease，续租lease，撤销lease

import etcd3
import json
import time
etcd = etcd3.client(host='127.0.0.1', port=2379)


# lease使用
# 创建新租约。返回lease对象
new_lease = etcd.lease(5)
print("new_lease:{}" .format(new_lease.__str__()))
etcd.get_lease_info(new_lease.id)
print("get_lease_info:{}" .format(etcd.get_lease_info(new_lease.id)))
# 如果租约到期，附加到该租约的所有密钥都将过期并删除。
# 可以发送租约保持活动消息以刷新 ttl。如果超时了的话续租仍然是可以成功的
for i in range(3):
    # 续租
    print("=========={}============" .format(i))
    res = list(etcd.refresh_lease(new_lease.id))
    # 打印续租是否成功信息
    time.sleep(3)
    print("refresh_lease resp:{}" .format(res))
# time.sleep(5) # 超时的话，打印error
# 撤销租约
try:
    res = etcd.revoke_lease(new_lease.id)
    print("revoke_lease resp:{}" .format(res.__str__()))
except:
    print("revoke_lease error")

# 打印信息如下：
'''
new_lease:
get_lease_info:header {
  cluster_id: 17237436991929493444
  member_id: 9372538179322589801
  revision: 2
  raft_term: 4
}
ID: 3632582335366158369
TTL: 4
grantedTTL: 5

==========0============
refresh_lease resp:[header {
  cluster_id: 17237436991929493444
  member_id: 9372538179322589801
  revision: 2
  raft_term: 4
}
ID: 3632582335366158369
TTL: 5
]
==========1============
refresh_lease resp:[header {
  cluster_id: 17237436991929493444
  member_id: 9372538179322589801
  revision: 2
  raft_term: 4
}
ID: 3632582335366158369
TTL: 5
]
==========2============
refresh_lease resp:[header {
  cluster_id: 17237436991929493444
  member_id: 9372538179322589801
  revision: 2
  raft_term: 4
}
ID: 3632582335366158369
TTL: 5
]
revoke_lease resp:None
'''

2、将lease attach到key上

import etcd3
import json
import time
etcd = etcd3.client(host='127.0.0.1', port=2379)

# 基础操作 put get delete
# 获取当前所有存储的key
print(list(etcd.get_all()))

# 创建新租约。返回lease对象
new_lease = etcd.lease(5)
print("new_lease:{}" .format(new_lease.__str__()))
etcd.get_lease_info(new_lease.id)
print("get_lease_info:{}" .format(etcd.get_lease_info(new_lease.id)))
# 将lease attach 到某个key上
etcd.put('/demo/key', 'hello world',lease=new_lease)
# 查看key是否存在
print(list(etcd.get_all()))
time.sleep(6)
# 查看key是否存在
print(list(etcd.get_all()))
etcd.close()

# 打印结果如下：
# [(b'hello', )]
# new_lease:
# get_lease_info:header {
#   cluster_id: 17237436991929493444
#   member_id: 9372538179322589801
#   revision: 2
#   raft_term: 4
# }
# ID: 3632582335366158376
# TTL: 4
# grantedTTL: 5

# [(b'hello world', ), (b'hello', )]
# [(b'hello', )]

watch使用

watch、watch_once

我们这里开一个进程watch某一个key，在另外一个进程里进行替换或者删除key。

import etcd3
import json
import time
etcd = etcd3.client(host='127.0.0.1', port=2379)

# 查看当前集群的kv
list_of_kv = list(etcd.get_all())
for value, kv_msg in list_of_kv:
    print("key:{}, value:{}" .format(kv_msg.key, value))
# key:b'key', value:b'hello'

# 对一个key进行watch
# 这里是阻塞的，我们在其他的程序中进行key的修改
events_iterator, cancel = etcd.watch('key')
for event in events_iterator:
    print(event)
print(cancel)
etcd.close()

import etcd3
import json
etcd = etcd3.client(host='127.0.0.1', port=2379)

# 修改key的value
print(etcd.replace('key', initial_value='hello', new_value='hello2'))
print(etcd.get('key'))
# 删除key
print(etcd.delete('key'))
print(etcd.get('key'))
etcd.close()

replace

delete

可以看到在接收到事件后watch进程仍在监听状态

我们可以使用watch_once在第一个事件后停止

watch_prefix、watch_prefix_once

监视一系列key

import etcd3
import json
import time
etcd = etcd3.client(host='127.0.0.1', port=2379)

print(etcd.put('demo/bar1', 'doot'))
print(etcd.put('demo/bar2', 'doot'))
print(etcd.put('demo/bar3', 'doot'))
print(etcd.put('demo/bar4', 'doot'))
# 查看当前集群的kv
list_of_kv = list(etcd.get_all())
for value, kv_msg in list_of_kv:
    print("key:{}, value:{}" .format(kv_msg.key, value))
# key:b'key', value:b'hello'

# 对一个key进行watch
# 这里是阻塞的，我们在其他的程序中进行key的修改
events_iterator, cancel = etcd.watch_prefix('demo/')
for event in events_iterator:
    print(event)
print(cancel)
etcd.close()

cancel_watch

当我们不想继续watch后，我们可以显式取消watch

这里我们在第3次处理watch事件后进行取消

import etcd3
import json
import time
etcd = etcd3.client(host='127.0.0.1', port=2379)

# 查看当前集群的kv
list_of_kv = list(etcd.get_all())
for value, kv_msg in list_of_kv:
    print("key:{}, value:{}" .format(kv_msg.key, value))
# key:b'key', value:b'hello'

# 对一个key进行watch
# 这里是阻塞的，我们在其他的程序中进行key的修改
cnt = 0
events_iterator, cancel = etcd.watch('key')
for event in events_iterator:
    cnt += 1
    print(event)
    print("handle times:{}" .format(cnt))
    if 3 == cnt:
        cancel() # cancel底层调用的就是cancel_watch(watchid)
        print("cancel watch")
etcd.close()

很显然，如果我们在主线程进行event监控，会造成主线程阻塞。

add_watch_callback

我们可以向watch加入回调函数，这样就不会阻塞主进程了。

import etcd3
import time
etcd = etcd3.client(host='127.0.0.1', port=2379)

def callback(resp):
    # resp is a events_iterator
    for event in resp.events:
        print("Key:{}发生改变, 新的value是:{}" .format(event.key, event.value))

etcd.add_watch_callback('key', callback)
# 程序主流程
while True:
    time.sleep(1)

参考

https://www.modb.pro/db/53046