FAQs

How secure is UMem Redis?

Access Security

UMem Redis is 'only accessible via the internal network and isolated by account', therefore only UHosts under the same account can access the UMem Redis instances.

Data Security

All memory data is persisted to the disk, preventing data loss due to service failure restarts, and avoiding the need for data to be preheated again.

What is the maximum capacity supported by a single instance?

The capacity limit of the master-replica Redis instance is 64GB

The distributed version of Redis theoretically has no capacity limit, but a work order is required for capacities exceeding 16TB.

Which protocols does the master-replica Redis support? Are they native protocols?

The master-replica Redis supports all native data read and write commands, but for security reasons, the following commands are disabled:

BGREWRITEAOF BGSAVE DEBUG CONFIG SAVE SHUTDOWN REPLICAOF FLUSHALL FLUSHDB

In addition, Users can flush data through the "Clear Data" function on the instance details page in the control panel, performing FLUSHALL or FLUSHDB to clear data.

What are the limitations of using the distributed version of Redis?

Apart from some unsupported protocols, a distributed Redis instance only supports one DB, meaning you can only select 0, and selecting others is meaningless. The keys command in Redis is performance-intensive, so try to reduce the frequency of keys usage in your business, or replace it with other methods. For batch commands or pipeline methods such as mget, mset, del, it is recommended not to exceed 1000 in batch quantity, as too many can easily cause request delay or timeout.

How does UMem Redis ensure the high availability of storage services?

UMem Redis instances have two storage nodes, master and replica, and they synchronize in real time to ensure data consistency. If the master node goes down, the system will automatically switch to the replica node to continue providing read and write services.

What to do if the storage capacity of UMem Redis is insufficient?

If the storage capacity is insufficient, it is recommended to expand it. If the master-replica version of Redis reaches the capacity limit (32GB or 64GB), it needs to be migrated to the NVMe version of the master-replica version of Redis or the distributed version of Redis. You can create a new instance of the corresponding specification, use UDTS to migrate by yourself, or contact technical support for migration; if the backend shards of the distributed version of Redis reach the capacity limit, you need to add shard splitting tasks for splitting and expansion.

If the expansion is not done in time, it may cause write failure (Note: The allkeys-lru elimination strategy is to eliminate data when writing, and there may also be write failures). Please expand in time when you receive alerts and reminders to avoid affecting the business.

What should I pay attention to for the high availability of the master-replica version of Redis?

The high availability of the master-replica version of Redis is based on the native Redis master-replica (replicaof) implementation. Because Redis's master-replica is asynchronous, URedis will automatically switch in the event of a failure; although the time is extremely short, theoretically, there may still be data differences.

What is the AOF file rewriting mechanism of the distributed version of Redis?

The AOF rewriting mechanism of each shard of the distributed version of Redis is consistent with the AOF rewriting mechanism of the master-replica version of Redis.

How many access IPs does the Master-replica Redis provide?

Master-replica Redis only provides one accessible IP. This IP will automatically migrate in the event of a failure; please do not use the IP of the replica, as it may become invalid when a failure occurs.

Does the replica node (replica) of the Master-replica Redis keep up to date with its master node (Master)? Updates to the master node (Master) are automatically replicated to its associated replica node (replica). However, due to Redis's asynchronous replication technology, for various reasons, the updates on the replica node may lag behind its Master node. Possible reasons include the I/O write volume of the Master node exceeding the synchronization speed of the replica node; or there is network latency between the Master node and the replica node. Therefore, there may be a lag or a certain degree of data inconsistency between the replica node and its Master node at some point.

What is the rule for deleting expired key data in UMem Redis?

UMem Redis has two ways to delete expired keys:

1. Active expiration, the system background will periodically detect and delete the keys when they are found to have expired.

2. Passive expiration, when a user accesses a key, if the key has expired, it will be deleted.

What is the default data eviction policy for UMem Redis?

The default eviction policy for distributed Redis is no-eviction.

The default eviction policy for Master-replica Redis is no-eviction. Users can change this in the file management of the control panel;

volatile-lru: The LRU algorithm is used to evict data from the set of data that has set an expiration time.

volatile-ttl: Select data that is about to expire from the set of data that has set an expiration time for eviction.

volatile-random: Randomly select data for eviction from the set of data that has set an expiration time.

allkeys-lru: The LRU algorithm is used to evict data from all data sets.

allkeys-random: Randomly select data for eviction from the data set.

no-eviction: Prohibition of data eviction.

Why doesn't the memory usage drop significantly after a large number of keys are deleted in the Master-replica Redis?

The memory usage of the Master-replica Redis is determined by taking the max of the used_memory (data size) and used_memory_rss (physical memory size) information returned by the Redis info command. In many cases, even if a large number of Keys are deleted, used_memory_rss may not decrease significantly, but used_memory will decrease accordingly. This is a phenomenon caused by Redis's memory management strategy. The ratio of used_memory_rss to used_memory is called the memory fragmentation rate, which is generally below 150%. If you want to reduce the fragmentation rate, you can consider running online fragmentation on the console, or restarting Redis, or migrating Redis in a non-standard way.

Does Redis expansion affect online services?

Upgrading or downgrading the capacity of the distributed and master-replica Redis on the console may involve data migration. If migration is needed, the console will prompt. The specific impact is as follows: Redis service is still available during the upgrade and downgrade period, but the load will increase when data synchronization starts, and there will be a flash cut of about 3 seconds when the high-availability IP switches. Please try to execute during the off-peak period of the business.

After the distributed Redis is expanded, the background may increase its shards to enhance processing capacity and improve performance. Adding shards is achieved through online migration. During the migration process, there may be a phenomenon of increased request latency, so it is generally arranged to perform migration operations in the early morning. If the migration volume is large, it will be done in the early morning multiple times. Depending on the actual data situation, the migration process may last for one or more early mornings to complete the entire migration task. If you need to migrate immediately, you can notify technical support.

How is Redis usage and usage rate calculated?

Usage: The maximum value taken from the used_memory (data size) and used_memory_rss (physical memory size occupied, i.e., how much memory is applied for use from the OS, there may be memory fragments in actual use) information returned by the info command is the usage.

Usage rate: Usage / Purchased capacity * 100%

What should be noted when restarting the master-replica version of Redis?

The restart process involves the management service calling the shutdown command on the Redis process to close Redis, and then pulling up the Redis process; after the Redis process is pulled up, Redis will load the AOF file data. Currently, for Redis with a capacity greater than 2G, the AOF file is generally large and the loading time may be relatively long. If the AOF persistence function is turned off in the configuration management, the data will be cleared after the restart, please operate with caution;

What is the AOF rewrite mechanism of the master-replica version of Redis?

What is the maximum number of connections for creating a new master-replica version of Redis?

Will backups be deleted after the Redis instance is deleted?

After the master-replica Redis instance is deleted, backups (including automatic backups and manual backups) will be retained for 7 days, and will be automatically recycled after 7 days.