Steem Pressure #7: Go Fork Yourself. Step by step guide to building and setting up a MIRA powered HF21-ready Steem consensus node.

avatar

The date of the next, 21st HardFork, has been set, and it is approaching quickly.

There's plenty of info around the web on how to build a Steem node from scratch.
I also wrote a guide some time ago: Steem Pressure #3: Steem Node 101.
But the trouble is that such tutorials are getting old and obsolete.

My previous one was made for version 19 at its early development stages.
Now, we have version 21 released, so it is a good opportunity for an update.


Video created for Steem Pressure series.

Hardware

Nothing fancy.
Actually we will use something reasonably modest and cheap.

Meet the mirabelle:
Intel Xeon CPU W3530 @2.80GHz
16GB DDR3 ECC 1333MHz RAM
2x 240GB SSD

Operating system

Ubuntu 18.04 LTS

Clean, basic server install.
No fancy tweaks or customizations.
/home is RAID0 because we need speed and space.
In the case of a failure, we just use a new one.

Let's Go

Few steps needs to be done as root:

Make sure that your system is up to date:

apt update && apt upgrade

Use ntp or chrony to keep your system clock in sync:

apt install ntp

Create an account for steem purposes:

useradd -s /bin/bash -m steem

Set appropriate values for open files limits needed by MIRA:
cat << 'EOF' >> /etc/security/limits.conf
steem soft nofile 65535 
steem hard nofile 65535
EOF

limits-conf.gif

Install packages needed for building Steem:
apt install \
  automake \
  build-essential \
  cmake \
  doxygen \
  libbz2-dev \
  libboost-date-time-dev \
  libboost-chrono-dev \
  libboost-context-dev \
  libboost-coroutine-dev \
  libboost-filesystem-dev \
  libboost-iostreams-dev \
  libboost-locale-dev \
  libboost-program-options-dev \
  libboost-serialization-dev \
  libboost-system-dev \
  libboost-test-dev \
  libboost-thread-dev \
  libreadline-dev \
  libsnappy-dev \
  libssl-dev \
  libtool \
  ncurses-dev
Now we continue as a user steem:

su - steem

Clone the steem repository:

git clone https://github.com/steemit/steem

Checkout the latest release:

cd steem
git checkout v0.21.0

Create build directory:

mkdir ~/build

Configure steem build:

cd ~/build

cmake -DCMAKE_BUILD_TYPE=Release \
  -DLOW_MEMORY_NODE=ON \
  -DCLEAR_VOTES=ON \
  -DSKIP_BY_TX_ID=OFF \
  -DENABLE_MIRA=ON \
  -DSTEEM_STATIC_BUILD=ON \
  ../steem

git-checkout-cmake.gif

Build steemd:

make -j4 steemd

build-steemd.gif

Build cli_wallet:

make -j4 cli_wallet

Create local bin directory for convenience:

mkdir ~/bin

Copy steemd and cli_wallet binaries to local bin directory:

cp -v ~/build/programs/steemd/steemd ~/bin
cp -v ~/build/programs/cli_wallet/cli_wallet ~/bin

Congratulations!

~/bin/steemd --version

Now you have required binaries.
Time to configure and run your node.

Create proper directory tree:

mkdir -pv ~/.steemd/blockchain

Create minimalistic config file for consensus node:
cat << 'EOF' >> ~/.steemd/config.ini
plugin = witness
plugin = condenser_api network_broadcast_api block_api
webserver-http-endpoint = 127.0.0.1:8090
webserver-ws-endpoint = 127.0.0.1:8090
EOF

config-ini.gif

Yes, you really don't need more than that for a very basic node defaults are OK.

To speed up reaching the head block:

Download blocks from a trusted source:
wget https://gtg.steem.house/get/blockchain/block_log \
  -O ~/.steemd/blockchain/block_log

block_log.gif

Now run:

~/bin/steemd --replay
and have fun!

This configuration is expected to reach the head block within 72 hours.

Footnotes:

  • You might want to run above inside screen.
  • Once it replays all the blocks from block_log, it will sync with Steem p2p network up to the head block and continue running.
  • Next time start steemd without --replay. You need to do that only if your changes will affect the state (such as adding or re-configuring plugins).
  • You might want to add account_by_key (for cli_wallet’s list_my_accounts feature) and account_history (to track your own account history) but in this guide I wanted to focus on a simplest case.
  • You need to wait for replay to complete before you can use API endpoint.

Previous episodes of Steem Pressure series

Introducing: Steem Pressure #1
Steem Pressure #2 - Toys for Boys and Girls
Steem Pressure #3 - Steem Node 101
Steem Pressure: The Movie ;-)
Steem Pressure #4 - Need for Speed
Steem Pressure #5 - Run, Block, Run!
Steem Pressure #6 - MIRA: YMMV, RTFM, TLDR: LGTM
Stay tuned for next episodes of Steem Pressure :-)


If you believe I can be of value to Steem, please vote for me (gtg) as a witness on Steemit's Witnesses List or set (gtg) as a proxy that will vote for witnesses for you.
Your vote does matter!
You can contact me directly on steem.chat, as Gandalf



Steem On

steem-pressure steem dev witness-category palnet
0
0
0.000
53 comments
avatar

1333MHz RAM 😲

0
0
0.000
Reply
avatar

something reasonably modest and cheap

:-)

0
0
0.000
Reply
avatar

very! indeed.... I could get tons of it... but I would not have power enough to support it.

0
0
0.000
Reply
avatar

..sounds like being possible also for me (one day..)..thx..up..follow you..

0
0
0.000
Reply
avatar

What the heck is the purpose for a witness proxy?

Off-topic yes, but maybe an idea for you to do a witness episode?

Anyways, this was an awesome and super easy step-by step! Thanks!

0
0
0.000
Reply
avatar

Witness proxy is an account that you chose to vote for witnesses on your behalf.
It's meant for people that don't pay that much of attention to what's going on with witnesses and are comfortable trusting their witness proxy to make those choices for them.
So if you set someone to be your witness proxy, whenever they decide to approve or unapprove some witness, your votes will be exactly the same.

0
0
0.000
Reply
avatar

Great explanation! I love that you included a use case, thank you!

0
0
0.000
Reply
avatar

Thanks. I’ve been looking for an updated Steem Node setup guide. If I have a much more powerful machine with 128Gb RAM what can I add to make it a full node?

Posted using Partiko iOS

0
0
0.000
Reply
avatar

I wouldn't recommend using "full node" anymore, i.e. monolithic instance.
I would rather use fat node and low mem node of AH - your machine might be good for one of those. Adding Hivemind and Jussi there, will make it pretty nice set to satisfy full API needs.

0
0
0.000
Reply
avatar

Perfect timing 👨🏾‍🚀👍🏽

can get @reggaesteem node up asap

🍻!BEER

Posted using Partiko iOS

0
0
0.000
Reply
avatar

It's good to start with a public seed node and try to keep it up and running.
To make your seed node available to general public you will need to add:
p2p-endpoint = 0.0.0.0:2001
and let people know on a steem.chat's #seed-nodes channel that a new seed-node is in town ;-)
Static IP is required and some reliable hostname is preferred.
(above is true for public seed node)

0
0
0.000
Reply
avatar

Everybody is looking forward to the date and the kickstart of the 21 hardfork and I must say, majority people can't wait

0
0
0.000
Reply
avatar

True effects of HF21 will not be instant. It might take a month or two before people will adjust their habits.

0
0
0.000
Reply
avatar

So if a full-fledged system is to be made from the instructions you have given here, how much time would it take for it to be made available to the public and what about the security of the chain?

0
0
0.000
Reply
avatar
(Edited)

In this configuration node will not be available to the public in other means than just being a part of Steem p2p network. It's a consensus node. In this configuration it can serve you as a simple API node (with subset of supported API calls), you can use it for example to broadcast transactions to the network.

To turn it into a public seed node, you need to add:
p2p-endpoint = 0.0.0.0:2001
which will make it to listen on all interfaces on port 2001 (default p2p port for Steem)
and then let people know on a steem.chat's #seed-nodes channel that a new seed-node is in town ;-)
Static IP is required and some reliable hostname is obviously good to have.

It will take about 72 hours using that example hardware/software configuration to reach the head block.

When it finish replaying, its API endpoint will be available. You can use some proxy with SSL termination to make it available to the public too, but keep in mind that API is limited.

Such node might improve distribution of the system but it doesn't play the same role as in PoW networks.

0
0
0.000
Reply
avatar
(Edited)

Hoping this to change a bit on the next SMT model...

0
0
0.000
Reply
avatar

its not a standard thing...everyone will have its "crazy" things... I for example, would be more enterprise model... if I had the money to support it. Not that I could not do it with what I have (which I have done in the past)...

Everything depends... ask something more specific, and I might be able to shed some more light. If that's where you want to go.

0
0
0.000
Reply
avatar

Looking forward for an updated Steem Node setup guide !!!!!!

0
0
0.000
Reply
avatar

Maybe you should look backward then? ;-) That's the guide you are looking for.

0
0
0.000
Reply
avatar

Can I start one with dule core 8 gb ram and 500gb hard drive

Posted using Partiko Android

0
0
0.000
Reply
avatar

All depends on details. 8GB RAM itself wouldn't be that bad but you will experience much higher replay times, especially when your drive is a platter drive, i.e. not SSD.

0
0
0.000
Reply
avatar

Why do you think 8GB RAM would limit replay speed? When I was replaying my MIRA node, the RAM usage stayed below 3GB the entire time.

I used a platter HDD as well and my replay time was roughly 5 days.

0
0
0.000
Reply
avatar

Less for the cache / buffer. Are you using default database.cfg?

0
0
0.000
Reply
avatar

Not sure, didn't spend much time tweaking it since it didn't seem necessary for my setup.

{
  "global": {
    "shared_cache": {
      "capacity": "5368709120"
    },
    "write_buffer_manager": {
      "write_buffer_size": "1073741824"
    },
    "object_count": 62500,
    "statistics": false
  },
  "base": {
    "optimize_level_style_compaction": true,
    "increase_parallelism": true,
    "block_based_table_options": {
      "block_size": 8192,
      "cache_index_and_filter_blocks": true,
      "bloom_filter_policy": {
        "bits_per_key": 10,
        "use_block_based_builder": false
      }
    }
  }
}
0
0
0.000
Reply
avatar

Geesssss... coming back from the future... I need to reconsider my hardware... EOS made me go... 😣 with my set... Still waiting... when I am confortable, I will launch.

0
0
0.000
Reply
avatar

Congratulations @gtg!
Your post was mentioned in the Steem Hit Parade in the following category:

  • Pending payout - Ranked 3 with $ 80,84
0
0
0.000
Reply
avatar

Nice to see a new steem pressure post. I wonder how many new witnesses there will be and how many will stop over the next year.

Posted using Partiko Android

0
0
0.000
Reply
avatar

If I set up a node in this fashion, would I be able to execute RPC calls against it using Beem? I specifically believe I need the reward API from previous discussions.

I would guess I would not be needing condenser in that case. My intent is to test the SFR bot at some point and explore a mechanism so reward flags different coming from downvote mana.

Any thoughts / guidance would be appreciated.

Posted using Partiko Android

0
0
0.000
Reply
avatar

While you can use it to perform some calls, LOW_MEMORY_NODE might not be enough for you. Please note that rewards_api is for testing purpose only, and explicitly said that it's not meant for use in production.

0
0
0.000
Reply
avatar

I may be getting the API needed confused in that case.

Our bot does approximate flag rshares to SBD conversions using methods of the Beem Steem class for it's voting.

Looks like the method uses get_sbd_per_rshares which calls get_reward_fund. That uses the database API if I am understanding correctly.

That should be fine, right?

Posted using Partiko Android

0
0
0.000
Reply
avatar

The rewards_api is something brand new that was used to attempt to model the changes to rewards after HF21.

What you are referring to is found in condenser_api which the the api you will use the majority of the time. Sometimes there is the need for database_api but, neither of those are explicitly declared in the config.

0
0
0.000
Reply
avatar

I've been re-forking myself for over a week now, hoping to finish it up one century soon.

0
0
0.000
Reply
avatar

@gtg You have received a 100% upvote from @intro.bot because this post did not use any bidbots and you have not used bidbots in the last 30 days!

Upvoting this comment will help keep this service running.

0
0
0.000
Reply
avatar
(Edited)

I would use a hybrid RAID0 NVMe with RAID5/6 disks... but not recommended on Ubuntu... CentOS or RedHAT 100% ok! =) Tested. With snapshots backups daily and MANY recovery tests ...

0
0
0.000
Reply
avatar

Doesn't make much sense when it comes to simple consensus node.
That one is just cheap and simple.
What exactly is not recommended on Ubuntu, and why?

0
0
0.000
Reply
avatar
(Edited)

cached LVM filesystems... to be specific! Lots of bugs! Especially on converts and recovery situations... and if you don't use extX filesystems even worse...

0
0
0.000
Reply
avatar
(Edited)

For the record:
Replayed 35538728 blocks in 75h:51m:36s
Then it had to catch up 98913 blocks that were generated in the meantime.
That took another 31m:45s
So in total it took 76h:23m:21s

Also:
steem@mirabelle:~/.steemd/blockchain$ du -csh *

215G    block_log
273M    block_log.index
269M    rocksdb_account_authority_object
14M     rocksdb_account_history_object
14M     rocksdb_account_metadata_object
406M    rocksdb_account_object
32M     rocksdb_account_recovery_request_object
27M     rocksdb_block_summary_object
25M     rocksdb_change_recovery_account_request_object
14M     rocksdb_comment_content_object
20G     rocksdb_comment_object
263M    rocksdb_comment_vote_object
53M     rocksdb_convert_request_object
18M     rocksdb_decline_voting_rights_request_object
31M     rocksdb_dynamic_global_property_object
21M     rocksdb_escrow_object
24M     rocksdb_feed_history_object
13M     rocksdb_hardfork_property_object
75M     rocksdb_limit_order_object
19M     rocksdb_liquidity_reward_balance_object
17M     rocksdb_operation_object
39M     rocksdb_owner_authority_history_object
14M     rocksdb_pending_optional_action_object
14M     rocksdb_pending_required_action_object
27M     rocksdb_proposal_object
17M     rocksdb_proposal_vote_object
84M     rocksdb_rc_account_object
14M     rocksdb_rc_pool_object
11M     rocksdb_rc_resource_param_object
24M     rocksdb_reward_fund_object
70M     rocksdb_savings_withdraw_object
24M     rocksdb_transaction_object
47M     rocksdb_vesting_delegation_expiration_object
85M     rocksdb_vesting_delegation_object
35M     rocksdb_withdraw_vesting_route_object
14M     rocksdb_witness_custom_op_object
6.9G    rocksdb_witness_object
24M     rocksdb_witness_schedule_object
83M     rocksdb_witness_vote_object
244G    total
0
0
0.000
Reply
avatar

I would be interested to see benchmarks on the new AMD Epyc chips...

0
0
0.000
Reply
avatar

Unfortunately I don't have such hardware available. Ideally it will be same disks, similar RAM setup, and just a difference with CPU/MB vs some high performance Intel setup.

PS
Please take a look at your node, it's missing blocks and it will continue as there's no longer witness shutdown vop in place due to 51% escalation issue, so the only way currently to disable it, is operator nullifying signing key.

0
0
0.000
Reply
avatar

Oh, I didn't know that. I'll shut it down.

0
0
0.000
Reply
avatar

Better yet - come back! :-)

0
0
0.000
Reply
avatar

I have no time unfortunately.

0
0
0.000
Reply
avatar

Max open files on my node is 5mil.
Still the mira replay crashes at 5%.
Too many open files.
0.21.9 did run smoothly though:(

0
0
0.000
Reply
avatar

Could you please show some logs / configs?
Either here or on Steem.Chat
(and please be patient)

0
0
0.000
Reply
avatar

Did upgrade ulimit - n 65536
And doing currently a block log download and replay.
8 hours to go.
We will see..

I am now 4 days in, without a working witness node, I think I am patient. :)

0
0
0.000
Reply