350 people remotely: how it was

^{The administrator}

on ^duty 2 weeks ago, the level of paranoia in Tutu.ru was low, so we bought a pyrometer and a couple of accurate thermometers in the office, installed points with hand disinfection for surgeons in operating boxes at different ends, wiped all ventilation from the inside with a disinfectant and set up manic cleaning mode - disinfection of all surfaces every morning, in the usual cleaning cycle - disinfection of pens and the rest. Hazelnuts of nuts were removed in the kitchen and other things, they left only what was individually wrapped. It is fortunate that bananas and oranges grow immediately packaged. We conducted VPN exercises and worked out mechanisms for switching to a remote site.

Since Monday, the office has completely switched to remote work. About 350 people. For the last 15 years I have been appearing in my offices a couple of times a week, so the situation is familiar. Actually, this applies to almost anyone who travels a lot. But I'm not tied to infrastructure like call center software, repositories, and private segments with personal data. Since we are still a travel service, the rest in the office are prepared in much the same way, but with nuances.

70% of employees have office laptops. In the basic configuration, it is either Lenovo with an office dock, or a MacBook with a monitor into which everything else is stuck. That is, you can just tear off the car by default and take it home. And pick up your wireless keyboards and mice from the workplace. Naturally, not everyone has it - there are those who, for a number of reasons, want and love to sit at the desktop in the office.

Over the past week, the AXO has asked people to carry home devices for remote exercises, so on Monday we could just switch to the new mode.

Newly made remote employees without a laptop were divided into three categories:

• Your desktop at home. Ok, no question, you can work with it in common segments.
• Pick up the desktop from the office. Everything is clear too.
• Throw a bridge to a working desktop from home via remote access and work like this.

Another part was asked to give out laptops. We always have a reserve for new employees, plus a couple of weeks ago a batch of macbooks arrived (they must be ordered in advance, now there are supply disruptions and delays), so there were enough free devices. They remained now, but the reserve is already less. A couple of people asked for monitors, issued.

Before all this blizzard, the company already had remote admins and developers. The work of the administrator from home is no different from the usual work in the office. The entire Atlasian stack works great from anywhere, so no one in Jira knows where you are physically. The developers used a special VPN, which gave access to development services like repositories, analytics, and more. Admins are another VPN. OTP paid - there are licenses, in fact, for convenience. Among other features - dynamic ACLs, that is, the ability to give different users different accesses within the network. Now the licenses are over, so the admins have deployed a second VPN, which is already more cond in configuration, but solves the problem. Naturally,With the transition to mass VPN access, it was immediately necessary to add more sources to the information security monitoring and to work hard on the first false-positive incidents.

There are a few physical phones in the office. First of all, this means that at least one person must remain to monitor them. Secondly, they support call forwarding to mobiles, and the usual scenario is 6 beeps before transferring to mobile.

All closed circuits require a physical presence under the law. This means at least one personnel officer, security guard and lawyer from time to time in a deserted office.

There was still the administrator on duty. We have a server room for infrastructure tasks in the office (and prod in four geo-distributed data centers). Admin sometimes goes to the server room. He also monitors the issuance of laptops and desktops to those who return from vacation to this apocalyptic world. After a couple of weeks, constant duty will not be necessary.

Another extremely unpleasant moment for visiting the office is working with the originals of paper documents. Yes, now is 2020, but they still remained in a number of cases.

For all those who need an office for business needs, we do not recommend traveling by public transport. Taxi is paid by the company. You can choose any aggregator, then send the bookkeeping check - almost the same mechanics of the advance report work as on business trips.

Considering that in the company of very often traveling people there is always a remote site in one form or another, all protocols have long been worked out. No trackers that record time or take screenshots - only the result of tasks is always evaluated. How to control people in such a situation? No way. Need to follow the work. But there is an understanding that if someone fails in the jungle of procrastination and regularly creates “gags” for his team, then after everyone returns to the office, the possibility of remote work for such a person will be greatly limited at the maximum interval.

We have long been accustomed to communications through Zoom. It is very good for meetings, only before we brought one or two participants to the wall of the meeting room, and now we all meet in it. A free account is enough for most tasks - but today there will be a meeting for about 200 people online (monthly general demo), it will be more difficult there. At night there were test runs, in the afternoon an attempt to collect.

The usual communication takes place in a telegram most often, some teams use slack or our corporate Zulip (this is critical for the HR segment and accounting, for example). You can use Google Hangouts, Discord (it is convenient to create a permanent voice channel and adjust the volume for each participant separately) and even, sorry, Skype.

On the other hand, everything “kicked and found out” flew out of the processes. The pace of development will not slow down, but information about what happens in neighboring teams will go away. All the internal guilds went down the forest. All teams do not recruit new people (because at least a week before being removed is recommended in the office). More precisely, they are gaining, but so far there have been no cases of withdrawal to the team these days. This process has been slightly suspended, but soon we’ll come up with something.

Etiquette (we took part and transferred to ourselves from the book “Remote: an office is not required” by Jason Freud and David Hensson):

• It is advisable for all teams to be stable in contact from 11 to 17 Moscow time on weekdays, or to make sure that the absence of a person will not hurt someone.
• Sprints become as weekly as possible.
• , . , .
• - — .
• - - , . .
• — 30 . , , . , .
• 10 , , , . . . , . .
• — (), — . . «!», «! , , -».
• , , .

During the first couple of days of the mass exodus to the remote site, due to the large number of chats and people, several more good practices appeared in them:

• Agreements, logs of stand-ups and resumes of calls are denoted by stars in Zulip: when someone returns after a long absence in the channel, it is enough to read only the marked messages so as not to miss anything.
• So that the team members, getting out of the code, do not catch 100-200 unread ones, you need to call up if you can’t agree immediately. Conditionally, if a question takes more than 20 messages - a call and fixing in a chat. Voice channels in Discord proved to be excellent for constant discussion of one task, so as not to clutter up the team’s general chat. Flood to flood channels (separate from working team chats).
• No need to smear tasks at different entry points: it is better that they come in one particular channel.

The general paradigm of the remote is to divide the work into completely autonomous components. Each team should be full-stack and be responsible for creating the product from idea to implementation.

We don’t know how to make a general demo, but we will try soon. We don’t know how to keep the general context in development, but we’ll try to do the Monday 10-minute common stand-up in writing as a digest with the goals of each team for a week in confluence. We don’t know what is happening with the guilds, but now there will already be the first digests with what issues will be discussed, instructions will be updated on how to connect to guild events and channels.

We were lucky in that there was not much technical debt on this part of the work (in particular, on workflow stacks, knowledge bases, trackers and accesses), and we did everything right away and without panic. The HR department quickly switched to providing logistics transition and informing everyone on time about the details. Admins will survive any end of the world. Most of the external events and conferences were canceled by the organizers themselves. In general, everything is calm in Baghdad.