github.com/outbrain/consul@v1.4.5/website/source/docs/faq.html.md (about)

     1  ---
     2  layout: "docs"
     3  page_title: "Frequently Asked Questions"
     4  sidebar_current: "docs-faq"
     5  ---
     6  
     7  # Frequently Asked Questions
     8  
     9  ## Q: What is Checkpoint? / Does Consul call home?
    10  
    11  Consul makes use of a HashiCorp service called [Checkpoint](http://checkpoint.hashicorp.com)
    12  which is used to check for updates and critical security bulletins.
    13  Only anonymous information, which cannot be used to identify the user or host, is
    14  sent to Checkpoint . An anonymous ID is sent which helps de-duplicate warning messages.
    15  This anonymous ID can be disabled. In fact, using the Checkpoint service is optional
    16  and can be disabled.
    17  
    18  See [`disable_anonymous_signature`](/docs/agent/options.html#disable_anonymous_signature)
    19  and [`disable_update_check`](/docs/agent/options.html#disable_update_check).
    20  
    21  ## Q: Does Consul rely on UDP Broadcast or Multicast?
    22  
    23  Consul uses the [Serf](https://www.serf.io) gossip protocol which relies on
    24  TCP and UDP unicast. Broadcast and Multicast are rarely available in a multi-tenant
    25  or cloud network environment. For that reason, Consul and Serf were both
    26  designed to avoid any dependence on those capabilities.
    27  
    28  ## Q: Is Consul eventually or strongly consistent?
    29  
    30  Consul has two important subsystems, the service catalog and the gossip protocol.
    31  The service catalog stores all the nodes, service instances, health check data,
    32  ACLs, and KV information. It is strongly consistent, and replicated
    33  using the [consensus protocol](/docs/internals/consensus.html).
    34  
    35  The [gossip protocol](/docs/internals/gossip.html) is used to track which
    36  nodes are part of the cluster and to detect a node or agent failure. This information
    37  is eventually consistent by nature. When the servers detects a change in membership,
    38  or receive a health update, they update the service catalog appropriately.
    39  
    40  Because of this split, the answer to the question is subtle. Almost all client APIs
    41  interact with the service catalog and are strongly consistent. Updates to the
    42  catalog may come via the gossip protocol which is eventually consistent, meaning
    43  the current state of the catalog can lag behind until the state is reconciled.
    44  
    45  ## Q: Are _failed_ or _left_ nodes ever removed?
    46  
    47  To prevent an accumulation of dead nodes (nodes in either _failed_ or _left_
    48  states), Consul will automatically remove dead nodes out of the catalog. This
    49  process is called _reaping_. This is currently done on a configurable
    50  interval of 72 hours. Reaping is similar to leaving, causing all associated
    51  services to be deregistered. Changing the reap interval for aesthetic
    52  reasons to trim the number of _failed_ or _left_ nodes is not advised (nodes
    53  in the _failed_ or _left_ state do not cause any additional burden on
    54  Consul).
    55  
    56  ## Q: Does Consul support delta updates for watchers or blocking queries?
    57  
    58  Consul does not currently support sending a delta or a change only response
    59  to a watcher or a blocking query. The API simply allows for an edge-trigger
    60  return with the full result. A client should keep the results of their last
    61  read and compute the delta client side.
    62  
    63  By design, Consul offloads this to clients instead of attempting to support
    64  the delta calculation. This avoids expensive state maintenance on the servers
    65  as well as race conditions between data updates and watch registrations.
    66  
    67  ## Q: What network ports does Consul use?
    68  
    69  The [Ports Used](https://www.consul.io/docs/agent/options.html#ports) section of the Configuration documentation lists all ports that Consul uses.
    70  
    71  ## Q: Does Consul require certain user process resource limits?
    72  
    73  There should be only a small number of open file descriptors required for a
    74  Consul client agent. The gossip layers perform transient connections with
    75  other nodes, each connection to the client agent (such as for a blocking
    76  query) will open a connection, and there will typically be connections to one
    77  of the Consul servers. A small number of file descriptors are also required
    78  for watch handlers, health checks, log files, and so on.
    79  
    80  For a Consul server agent, you should plan on the above requirements and
    81  an additional incoming connection from each of the nodes in the cluster. This
    82  should not be the common case, but in the worst case if there is a problem
    83  with the other servers you would expect the other client agents to all
    84  connect to a single server and so preparation for this possibility is helpful.
    85  
    86  The default ulimits are usually sufficient for Consul, but you should closely
    87  scrutinize your own environment's specific needs and identify the root cause
    88  of any excessive resource utilization before arbitrarily increasing the limits.
    89  
    90  ## Q: What is the per-key value size limitation for Consul's key/value store?
    91  
    92  The limit on a key's value size is 512KB. This is strictly enforced and an
    93  HTTP 413 status will be returned to any client that attempts to store more
    94  than that limit in a value. It should be noted that the Consul key/value store
    95  is not designed to be used as a general purpose database. See
    96  [Server Performance](/docs/guides/performance.html) for more details.
    97  
    98  ## Q: What data is replicated between Consul datacenters?
    99  
   100  In general, data is not replicated between different Consul datacenters. When a
   101  request is made for a resource in another datacenter, the local Consul servers forward
   102  an RPC request to the remote Consul servers for that resource and return the results.
   103  If the remote datacenter is not available, then those resources will also not be
   104  available, but that won't otherwise affect the local datacenter. There are some special
   105  situations where a limited subset of data can be replicated, such as with Consul's built-in
   106  [ACL replication](/docs/guides/acl.html#outages-and-acl-replication) capability, or
   107  external tools like [consul-replicate](https://github.com/hashicorp/consul-replicate).